Battery health state estimation method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of battery management, in particular to a battery health state estimation method and device, computer equipment and a storage medium, and the battery health state estimation method comprises the steps: carrying out the difference analysis of a current capacity increment curve and an initial capacity increment curve of a battery through a DTW algorithm; the difference is converted into a visual image; and extracting the features of the image by using a convolutional neural network, thereby realizing the high-precision estimation of the SOH of the battery. According to the method, the time sequence data is converted into the image data, so that the convolutional neural network model can efficiently extract the nonlinear characteristics of the battery health state, and the SOH estimation precision is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a battery health state estimation method, device, computer equipment and storage medium. Background Art

[0002] Lithium-ion batteries have been widely used in transportation due to their long life, high specific power, and high energy. However, safety issues caused by inaccurate battery health state estimation and prediction have attracted widespread attention in academia.

[0003] Since the health status of lithium batteries is difficult to measure, has strong time variability, and is prone to interference (irreversibility and high nonlinearity), high-precision health status estimation and prediction are core technical challenges that need to be addressed urgently.

[0004] The measurement difficulty stems from the fact that the SOH (State of Health) of lithium-ion batteries is an internal parameter of the battery and cannot be directly measured using sensors. Relevant parameters such as voltage, current, and temperature can only be obtained through integration and approximation, which increases the difficulty of accurately identifying SOH parameters.

[0005] Strong temporal variability means that the SOH of lithium-ion batteries is not only closely related to environmental stresses such as temperature, current and load mode, but is also affected by the coupling of internal multiple parameters such as the internal ion motion state of the current, the ratio of positive and negative electrode active materials, and the intensity of electrochemical reactions.

[0006] Irreversibility refers to the fact that most current prediction methods use irreversible offline data for prediction, which is affected by individual differences and has low repeatability. Furthermore, online recognition is less reliable and timely.

[0007] High nonlinearity means that during actual operation, the health status of lithium-ion batteries is cross-coupled by multiple internal and external factors, and its degradation curve is highly nonlinear, which poses a challenge to the accurate identification and reliability of SOH. Summary of the Invention

[0008] In view of this, the present invention provides a battery health state estimation method, apparatus, computer equipment and storage medium to improve the accuracy of battery health state estimation.

[0009] In a first aspect, the present invention provides a method for estimating the state of health of a battery, comprising the following steps: obtaining battery management parameters during the charge and discharge stages of the battery; determining a current capacity increment curve based on the battery management parameters; obtaining an initial capacity increment curve of the battery; calculating the difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping (DTW) algorithm, and converting the difference into an image; and inputting the image into a trained convolutional neural network model to obtain a first estimated value of the current state of health of the battery.

[0010] The present invention uses the DTW algorithm to perform a difference analysis on the current capacity increment curve and the initial capacity increment curve, and converts the difference information into a visual image. The image form can more intuitively reflect the evolution trend of the polarization characteristics of the battery at different aging stages. The convolutional neural network (CNN) model can efficiently extract the nonlinear characteristics of battery aging based on the image, thereby significantly improving the accuracy of SOH estimation. In addition, the DTW algorithm has good time series elasticity matching capabilities and can effectively cope with voltage curve deformation caused by temperature fluctuations, changes in charge and discharge rates, etc.; without relying on constant test conditions, it can still stably extract aging features, significantly enhancing the generalization performance and robustness of the model in real usage scenarios. Different from the related art method of directly using the original data set as CNN input, the present invention converts the DTW difference results into images as input. The image data structure is unified and the format is standardized, which is more suitable for CNN fast processing. Compared with traditional high-dimensional time series data analysis, image input greatly improves the model training and reasoning speed, and significantly saves computing resources.

[0011] In an optional embodiment, after obtaining the first estimated value of the current battery health state, the battery health state estimation method also includes the following steps: inputting the battery management parameters into a preset twin model to obtain a second estimated value of the current battery health state; using the second estimated value to correct the first estimated value to obtain a third estimated value of the current battery health state.

[0012] After obtaining the first estimated value of the current battery health status (based on convolutional neural network image recognition), the present invention further inputs the battery management parameters into the preset electrochemical twin model to obtain the second estimated value, and then corrects the two estimated results through a weighted fusion strategy to obtain the final third estimated value; this design that combines the model-based method (Model-based) with the data-driven method (Data-driven) effectively makes up for the shortcomings of a single method in modeling error, generalization ability or real-time performance; significantly reduces the bias and variance of SOH estimation, and improves the overall prediction accuracy. The aging feature matrix generated based on the DTW algorithm can capture the dynamic evolution of the battery differential voltage curve at different aging stages. The convolutional neural network model extracts spatial structural features from the image perspective and belongs to the macroscopic level of characterization; while the electrochemical twin model simulates the internal reaction kinetics of the battery from the perspective of microscopic physical mechanisms; the two form multi-scale information complementarity, which not only retains the flexibility of data-driven, but also enhances the physical interpretability of the model, thereby improving the robustness and environmental adaptability of the entire system.

[0013] In an optional embodiment, the battery management parameters are the current-time correspondence and the voltage-time correspondence during the charge and discharge stages. Determining the current capacity increment curve based on the battery management parameters includes the following steps: obtaining the capacity charged into the battery at each time point during the charge and discharge stages based on the current-time correspondence to obtain the capacity-time correspondence; obtaining a voltage-capacity curve based on the voltage-time correspondence and the capacity-time correspondence; and performing differential processing on the voltage-capacity to obtain the current capacity increment curve.

[0014] The present invention utilizes the current-time and voltage-time relationships during the battery's charge and discharge processes, and through mathematical modeling and data processing, derives the capacity-time relationship and further constructs a voltage-capacity curve. The voltage-capacity curve is then differentiated to obtain the current capacity increment curve (dQ / dV). The entire process is entirely based on existing monitoring data in the BMS system, eliminating the need for new sensors or other hardware equipment, resulting in excellent engineering feasibility and deployment cost advantages. The capacity increment curve can effectively reflect key physical and chemical processes such as phase change behavior, lithium ion insertion / deinsertion characteristics, and changes in electrode materials during the battery's charging process. Changes in the curve morphology can serve as an important basis for determining battery state of health (SOH), electrode matching, and the presence of internal short circuits. Compared to using only macroscopic parameters such as voltage and capacity, the capacity increment curve provides deeper information on aging mechanisms, helping to improve the accuracy and interpretability of SOH estimation. In addition, in the early stages of battery aging, the voltage platform may not have changed significantly, but the peak position, peak width, area and other characteristics of the capacity increment curve have already shifted; using the DTW algorithm to dynamically compare the initial capacity increment curve with the current curve can accurately capture early aging signals; providing reliable data support for early warning and life prediction of the battery health management system. Through detailed analysis of this curve, we can deeply understand the working principle of the battery and its changes over time.

[0015] In an optional embodiment, the difference between the current capacity increment curve and the initial capacity increment curve is calculated using the dynamic time warping algorithm, and the difference is converted into an image, which includes the following steps: using the current capacity increment curve as the query sequence Q of the dynamic time warping algorithm and the initial capacity increment curve as the template sequence C of the dynamic time warping algorithm; constructing a distance matrix D, where the current element D[i][j] in the distance matrix D represents Q i and C j The distance between them; create a cumulative distance matrix Y, where the current element Y[i][j] in the cumulative distance matrix Y is used to store the minimum cumulative distance from the starting point (0, 0) to the current point (i, j); use dynamic programming principles and path constraints to fill the cumulative distance matrix Y, where the path constraints include boundary conditions and window constraints, and the window constraints are used to pass through the range of the window size limit j in the process of filling the cumulative distance matrix Y; after the cumulative distance matrix Y is filled, trace back from the end point to the starting point, and obtain the optimal alignment path according to the direction of the minimum cumulative distance selected at each step; draw the image according to the optimal alignment path.

[0016] DTW was originally widely used in time series signal processing fields such as speech recognition and action recognition, and this invention introduced it into the field of battery health management for the first time, for dynamic matching and difference extraction between differential voltage curves (dv / dt), achieving deeper modeling from voltage amplitude to voltage change trend. In addition, the present invention also converts the optimal alignment path output by DTW into an image, forming a new way of expressing data. The image form is naturally suitable for feature extraction by convolutional neural networks (CNN), enabling the model to learn key information about battery aging directly from the spatial structure.

[0017] The present invention uses DTW to align the current capacity increment curve with the initial capacity increment curve, which can accurately capture the morphological shift and local deformation caused by aging. Compared with traditional methods based on absolute voltage values ​​or simple statistics, this method can better reflect the true health status of the battery. Traditional methods rely on constant charging and discharging conditions, while factors such as temperature and current rate change frequently in real usage scenarios. DTW has good elastic matching capabilities and can automatically compensate for the voltage curve deformation caused by these factors, so that aging features can be accurately extracted even under non-ideal test conditions, improving the practicality and generalization ability of the system; moreover, the DTW image is used as the model input to represent the polarization voltage gradient change in an image-based manner, which greatly simplifies the feature extraction process. It should be noted that although DTW itself has a high computational complexity, the present invention can greatly reduce the computational overhead through path constraints. After converting the differences into images, CNN can be further used for efficient feature extraction and compression; the overall process is more lightweight, fast, and resource-friendly than traditional high-dimensional feature engineering.

[0018] Furthermore, traditional DTW algorithms don't consider window constraints, meaning that during the alignment process, a point in one sequence can be aligned with any point in the other sequence. This can lead to an overly "loose" alignment, especially when the sequence lengths differ significantly. This invention adds window constraints, ensuring that for each i, j, the points can only be within the range of i±window_size (window size). This ensures that the points in alignment don't deviate too far, maintaining temporal consistency. It also limits the matching range of each point, forcing the alignment path to extend only within a limited region centered on the diagonal. This further improves the computational efficiency of the DTW algorithm, enhances its noise immunity, and avoids overfitting and invalid matches.

[0019] In an optional embodiment, drawing the image according to the optimal alignment path includes: obtaining an optimal distance matrix according to the optimal alignment path, and drawing a heat map according to the optimal distance matrix; or; drawing a heat map according to the optimal alignment path.

[0020] That is to say, a heat map can be drawn according to the optimal distance matrix, or a heat map can be drawn according to the optimal alignment path. It should be noted that drawing a heat map according to the optimal distance matrix is ​​a preferred solution. This is because the heat map drawn according to the optimal distance matrix shows the cumulative distances of all possible paths in the entire alignment process, reflects the global similarity structure between sequences, and can intuitively display the alignment difficulty of different regions (such as high-cost areas). In addition, the heat map drawn according to the optimal distance matrix can reveal multiple potential alignment paths (such as local minimum areas), helping to analyze whether there are multiple reasonable alignment methods.

[0021] In an optional embodiment, the method for determining the window size in the window constraint includes: determining the data volume according to the current capacity increment curve and the initial capacity increment curve; determining the window size according to the data volume, wherein the window size is positively correlated with the data volume.

[0022] In this way, the window size can be dynamically adjusted according to the amount of data, and the window size can be accurately determined to avoid overfitting and invalid matching. This not only improves the accuracy and stability of data processing, but also enhances the intelligent adjustment capability, making it suitable for dynamically changing data processing scenarios.

[0023] In an optional embodiment, the convolutional neural network model includes a convolution layer, a pooling layer, a flattening layer and a fully connected layer; the convolution layer is used to extract features of the image to obtain a feature map; the pooling layer is used to pool the feature map; the flattening layer is used to convert the pooled feature map into a one-dimensional vector; the fully connected layer is used to obtain a first estimated value based on the one-dimensional vector.

[0024] This convolutional neural network model, which includes convolutional layers, pooling layers, flattening layers, and fully connected layers, has demonstrated powerful feature extraction capabilities, good robustness, and efficient state estimation performance in the field of battery health status monitoring, providing strong support for the realization of intelligent battery management.

[0025] In an optional embodiment, the convolutional neural network model further includes a shielding layer, which is disposed before the convolutional layer and is used to preprocess the image to shield alignment paths in the image that are away from the diagonal line.

[0026] By masking alignment paths away from the diagonal, we can focus on the most important parts of the image—those likely to contain features highly relevant to battery health assessment. This approach helps increase the relevance of the data input to the convolutional layer, thereby improving the accuracy of subsequent feature extraction. Furthermore, masking unimportant areas effectively reduces the impact of noise, allowing the model to focus more on learning truly useful features, thereby improving its overall performance and robustness and reducing unnecessary computation.

[0027] In a second aspect, the present invention also provides a battery health status estimation device, comprising a first acquisition module, a current capacity increment curve determination module, a second acquisition module, an image determination module and a first battery health status determination module; wherein the first acquisition module is used to obtain battery management parameters during the battery charging and discharging stages; the current capacity increment curve determination module is used to determine the current capacity increment curve based on the battery management parameters; the second acquisition module is used to obtain the initial capacity increment curve of the battery; the image determination module is used to calculate the difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm, and convert the difference into an image; the first battery health status determination module is used to input the image into a trained convolutional neural network model to obtain a first estimated value of the current battery health status.

[0028] In a third aspect, the present invention also provides a computer device comprising a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the battery health status estimation method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0029] In a fourth aspect, the present invention further provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the battery health status estimation method of the first aspect or any corresponding embodiment thereof.

[0030] In a fifth aspect, the present invention further provides a computer program product, comprising computer instructions, which are used to enable a computer to execute the battery health status estimation method of the above-mentioned first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 is a flowchart of a method for estimating a battery health state according to an embodiment of the present invention;

[0033] Figure 2 is a flowchart of another battery health status estimation method according to an embodiment of the present invention;

[0034] Figure 3 is a schematic diagram of an example of a voltage-capacity curve according to an embodiment of the present invention;

[0035] Figure 4 is a schematic diagram of an example of a capacity increment curve according to an embodiment of the present invention;

[0036] Figure 5 dQ / dV curves showing changes in dQ / dV with voltage when the battery health status is 100% and 86.1% respectively in an embodiment of the present invention;

[0037] Figure 6 DTW difference heatmap with window constraint when the battery health status is 100% and 86.1% respectively according to an embodiment of the present invention;

[0038] Figure 7 is a flowchart of a battery health status estimation method according to an embodiment of the present invention;

[0039] Figure 8 1 is a flow chart of a method for estimating the SOH of a power battery based on a battery twin model and CNN according to an embodiment of the present invention;

[0040] Figure 9 is a schematic diagram of an example of a battery twin model according to an embodiment of the present invention;

[0041] Figure 10 is a structural block diagram of a battery health status estimation device according to an embodiment of the present invention;

[0042] Figure 11 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. 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 those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0044] According to an embodiment of the present invention, an embodiment of a battery health status estimation method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0045] This embodiment provides a battery health status estimation method, which can be used in the above-mentioned computer device. Figure 1 FIG. 1 is a flow chart of a method for estimating a battery health state according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0046] Step S101: obtaining battery management parameters during the battery charging and discharging phases.

[0047] Specifically, the vehicle's BMS (Battery Management System) system can be used to obtain battery management parameters during the battery charging and discharging stages of the vehicle in real time.

[0048] Step S102: determining a current capacity increment curve according to battery management parameters.

[0049] The current capacity increment curve of a battery refers to a curve drawn by analyzing the rate of change of capacity relative to voltage during the battery's charge and discharge stages, namely dQ / dV (where Q represents capacity and V represents voltage).

[0050] A battery's current capacity increment curve can reveal information about phase changes during charging, its health during charging, electrode matching, and any internal short circuits. It can also effectively reflect battery aging. The capacity increment curve is a crucial tool for studying battery performance, aging mechanisms, and fault diagnosis. Detailed analysis of this curve can provide a deep understanding of the battery's operating principles and how they change over time.

[0051] Step S103: obtaining the initial capacity increment curve of the battery.

[0052] Specifically, the initial capacity increment curve of the battery can be obtained by analyzing the voltage and capacity data of a new battery or during the first charge and discharge process after several cycles to ensure that the battery activation is complete.

[0053] Step S104: Calculate the difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm, and convert the difference into an image.

[0054] Dynamic Time Warping (DTW) is a method used to measure the similarity between two time series, even if they are not strictly aligned on the time axis. In battery charge and discharge curve analysis, it can also be used to compare the differences between the capacity increment curves of batteries at different stages.

[0055] The DTW algorithm is applied to calculate the distance between two sets of capacity increment curves. DTW outputs a cumulative distance matrix that represents the best matching path between the two time series and the overall degree of difference between them.

[0056] Specifically, the image can be a heat map. The cumulative distance matrix generated by the DTW algorithm can be used as the data source of the heat map. Each matrix element represents the local similarity or distance value between two curves at the corresponding position.

[0057] Specifically, you can use Python's Matplotlib or Seaborn library to draw a heat map.

[0058] Step S105: Input the image into the trained convolutional neural network model to obtain a first estimated value of the current battery health state.

[0059] That is, the input of the convolutional neural network model is an image, and the output is an estimated value of the battery health state. The use of the convolutional neural network model to estimate the SOH in this embodiment has the following advantages: (1) The convolutional neural network can automatically learn the features of the image without the need for manual design or selection of feature extractors. (2) The convolutional neural network can utilize the spatial structure information of the image to maintain the invariance of the image to translation, rotation, and scaling. (3) The convolutional neural network can construct a deep network structure by stacking multiple convolutional layers and pooling layers, thereby improving the expression and generalization capabilities of the model.

[0060] This embodiment uses the DTW algorithm to perform a difference analysis on the current capacity increment curve and the initial capacity increment curve, and converts the difference information into a visual image. The image form can more intuitively reflect the evolution trend of the polarization characteristics of the battery at different aging stages. The convolutional neural network (CNN) model can efficiently extract the nonlinear characteristics of battery aging based on the image, thereby significantly improving the accuracy of SOH estimation. In addition, the DTW algorithm has good time series elasticity matching capabilities and can effectively cope with voltage curve deformation caused by temperature fluctuations, changes in charge and discharge rates, etc.; without relying on constant test conditions, it can still stably extract aging features, significantly enhancing the generalization performance and robustness of the model in real usage scenarios. Unlike the method of directly using the original data set as CNN input in related technologies, the present invention converts the DTW difference results into images as input. The image data structure is unified and the format is standardized, which is more suitable for CNN fast processing. Compared with traditional high-dimensional time series data analysis, image input greatly improves the model training and inference speed, and significantly saves computing resources.

[0061] Compared to the discharge phase, the battery health status estimation method provided by the present invention is more accurate during the charging phase. Therefore, this embodiment uses the charging phase as an example to explain the battery health status estimation method in detail. The battery health status estimation method provided by this embodiment can be used in computer equipment. Figure 2 FIG. 1 is a flow chart of another method for estimating the state of health of a battery according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0062] Step S201: Obtain battery management parameters during the battery charging phase.

[0063] Specifically, the battery management parameters are the current-time correspondence and the voltage-time correspondence during the charging phase.

[0064] Step S202: Determine the current capacity increment curve according to the battery management parameters.

[0065] In an optional embodiment, determining the current capacity increment curve according to the battery management parameters includes the following steps:

[0066] Step S2021: obtaining the capacity charged into the battery at each time point in the charging phase according to the current-time correspondence relationship, and obtaining the capacity-time correspondence relationship.

[0067] During the battery charging process, a series of current-time data points (It) are usually collected. In order to calculate the total capacity Q(t) that has been charged in the battery at any time t, the ampere-hour integration method can be used. Specifically, the calculation formula is:

[0068] In practice, since the data is discretely sampled, numerical integration methods such as trapezoidal integration or rectangular integration are used for approximate calculations.

[0069]

[0070] In the above formula, I i represents the current value of the i-th sampling point; Δt i Represents the time interval between two adjacent sampling points; k represents the data point index corresponding to the current time.

[0071] In this way, a time-related capacity sequence Q(t) can be obtained, thereby drawing a capacity-time correspondence (Qt curve), which reflects the trend of battery capacity changing with time during the charging process.

[0072] Step S2022: Obtain a voltage-capacity curve according to the voltage-time correspondence relationship and the capacity-time correspondence relationship.

[0073] During the battery charging process, the collected battery management parameters include not only the current-time correspondence, but also the voltage-time correspondence (Vt). By making a one-to-one correspondence between the previously calculated capacity-time correspondence Q(t) and the voltage-time correspondence (Vt), a new two-dimensional curve can be constructed: the voltage-capacity curve (VQ curve), as shown in the following example: Figure 3 shown.

[0074] Specifically, for each time point t k , there is a corresponding voltage value V(t k ) and capacity value Q(t k ), so we can construct a two-dimensional graph with capacity as the horizontal axis and voltage as the vertical axis.

[0075] Step S2023: performing differentiation processing on the voltage-capacity curve to obtain a current capacity increment curve.

[0076] Specifically, the current capacity increment curve can be obtained using the following formula:

[0077]

[0078] Among them, N is the sum of the number of sampling points in a certain time period or a certain voltage interval; I is the current; f is the sampling frequency, that is, the number of samples per minute; ΔV is the interval width, that is, the voltage difference between two consecutive sampling points or the voltage change range within a certain analysis interval; ΔQ is the capacity difference between two consecutive sampling points; ΔT is the time difference between two consecutive sampling points; V2 represents the voltage of the latter sampling point in two consecutive sampling points; V1 represents the voltage of the former sampling point in two consecutive sampling points; Q2 represents the capacity of the latter sampling point in two consecutive sampling points; Q1 represents the capacity of the former sampling point in two consecutive sampling points. It should be noted that the sampling interval between two consecutive sampling points is very short, at most 0.1S, and the current can be regarded as unchanged. I can be the current of the former sampling point in two consecutive sampling points, or the current of the latter sampling point in two consecutive sampling points. For example, the voltage-capacity is differentiated to obtain the current differential voltage curve as shown below. Figure 4 shown.

[0079] This embodiment utilizes the current-time and voltage-time relationships during the battery charging process, deriving the capacity-time relationship through mathematical modeling and data processing. It then constructs a voltage-capacity curve, which is then differentiated to obtain the current capacity increment curve (dQ / dV). This entire process is based entirely on existing monitoring data in the BMS system, eliminating the need for new sensors or other hardware, resulting in excellent engineering feasibility and deployment cost advantages. The capacity increment curve effectively reflects key physical and chemical processes occurring during the battery charging process, including phase change behavior, lithium-ion insertion / deinsertion characteristics, and changes in electrode materials. Changes in the curve morphology serve as an important basis for determining battery state of health (SOH), electrode compatibility, and the presence of internal short circuits. Compared to using only macroscopic parameters such as voltage and capacity, the capacity increment curve provides deeper information on aging mechanisms, helping to improve the accuracy and interpretability of SOH estimation. In addition, in the early stages of battery aging, the voltage platform may not have changed significantly, but the peak position, peak width, area and other characteristics of the capacity increment curve have already shifted; using the DTW algorithm to dynamically compare the initial capacity increment curve with the current curve can accurately capture early aging signals; providing reliable data support for early warning and life prediction of the battery health management system. Through detailed analysis of this curve, we can deeply understand the working principle of the battery and its changes over time.

[0080] Figure 5 dQ / dV curves of the battery in the embodiment of the present invention are shown as follows: Figure 5 It can be seen that the dQ / dV curve has obvious differences in shape with different battery health states, indicating a strong mapping relationship with SOH.

[0081] Step S203: Obtaining the initial capacity increment curve of the battery.

[0082] The initial capacity increment curve is a test curve when the battery SOH is 100%.

[0083] Step S204: Calculate the difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm, and convert the difference into an image.

[0084] The dynamic time warping algorithm is an algorithm used to measure the distance between two time series. It matches the two time series by finding an optimal alignment path so that the cumulative distance between them is minimized.

[0085] An alignment path is a set of pairs of points found between two time series that represent corresponding elements in the two series. This path shows how to "warp" the time dimension to best align the two series. For example, given two time series Q and C, the alignment path will indicate which point in one series corresponds to which point in the other series.

[0086] In an optional embodiment, calculating the difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm and converting the difference into an image includes the following steps:

[0087] Step S2041: Use the current capacity increment curve as the query sequence Q of the dynamic time warping algorithm, and use the initial capacity increment curve as the template sequence C of the dynamic time warping algorithm.

[0088] Specifically, two time series are defined in the dynamic time warping algorithm, namely the query sequence and the template sequence, where the query sequence can be expressed as Q = Q1, Q2, ..., Q n , the template sequence can be expressed as C=C1,C2,…,Cm.

[0089] Step S2042: Construct a distance matrix D, where the current element D[i][j] in the distance matrix D represents Q i and C j The distance between them.

[0090] Specifically, we can construct a (m+1)×(n+1) matrix D, where n represents the length of the query sequence, m represents the length of the template sequence, D[1,1] is initialized to 0, and the rest of the positions are initialized to infinity. The matrix element D(i, j) represents Q i and C j The distance d(Q i , C j ), this distance can be calculated using the Euclidean distance, i.e. d(Q i ,C j )=|Q i -C j |.

[0091] Step S2043: Create a cumulative distance matrix Y, where the current element Y[i][j] in the cumulative distance matrix Y is used to store the minimum cumulative distance from the starting point (0, 0) to the current point (i, j).

[0092] Specifically, a cumulative distance matrix Y of the same size as the distance matrix D can be created to store the minimum cumulative distance from the starting point to the current point. The cumulative distance matrix Y is initialized so that Y[0][0]=D[0][0].

[0093] Step S2044: Fill the cumulative distance matrix Y using dynamic programming principles and path constraints.

[0094] The dynamic programming principle ensures that the alignment path can only move rightward, upward, or diagonally. Specifically, for each element Y[i][j] in the cumulative distance matrix, its value is calculated by the following formula:

[0095] Y[i][j]=D[i][j]+min(Y[i-1][j],Y[i][j-1],Y[i-1][j-1]), which means that the minimum cumulative distance to reach point (i,j) is equal to the distance to that point plus the minimum cumulative distance from the previous point (which can be above, left or upper left) to reach here, that is, if you are currently at (i,j), the next step can only reach (i+1,j), (i,j+1) or (i+1,j+1), which ensures that the time order will not be reversed.

[0096] In this embodiment, the path constraint conditions include boundary conditions and window constraints, wherein the boundary conditions mean that the path must start from the starting point of one sequence to the starting point of another sequence, and end from the end point of one sequence to the end point of another sequence.

[0097] Traditional DTW algorithms don't consider window constraints. This means that during the alignment process, a point in one sequence can be aligned with any point in the other sequence. This can lead to an overly loose alignment, especially when the sequence lengths differ significantly. To constrain this alignment, adding window constraints is necessary. Window constraints restrict a point in one sequence to being aligned with points in the other sequence that are relatively close in position, thus maintaining temporal consistency and avoiding excessive distortion.

[0098] Window constraints limit the range of j by using the window size during the traversal process. Specifically, for each i, j can only be within the range of i ± window_size. This ensures that the points during alignment do not deviate too far, maintaining temporal consistency. It also limits the matching range of each point, forcing the alignment path to extend only within a limited area centered on the diagonal line. This further improves the computational efficiency of the DTW algorithm, enhances its noise immunity, and avoids overfitting and invalid matching.

[0099] Specifically, the method for determining the window size in the window constraint includes: determining the data volume based on the current capacity increment curve and the initial capacity increment curve; and determining the window size based on the data volume. For example, the intersection of the current capacity increment curve and the initial capacity increment curve can be used as the data volume, and the window size can be obtained by using the difference between the data volume and the preset ratio value. When the preset ratio value is 8%-20%, it can avoid the window being too small to cause alignment, and it can also avoid the window being too large to reduce the effect of the window constraint. In this way, the window size can be dynamically adjusted according to the data volume, the window size can be accurately determined, and overfitting and invalid matching can be avoided. This not only improves the accuracy and stability of data processing, but also enhances the intelligent adjustment capability, making it suitable for dynamically changing data processing scenarios.

[0100] Figure 6 It is a DTW difference heat map with window constraint when the battery health status is 100% and 86.1% respectively according to an embodiment of the present invention. Figure 6 Due to the window limitation, the reddish-brown part is not explored. In the exploration area, the more blue the color is, the smaller the difference between the two curves in this area is. As the color changes from blue to red, the difference between the two curves gradually increases. Figure 6 The difference between the current dQ / dV curve and the original dQ / dV curve is the largest when they are aligned in the sampling point interval [30-55], and the corresponding voltage interval is 3.86V-4.0V.

[0101] Step S2045: After the cumulative distance matrix Y is filled, trace back from the end point to the starting point, and obtain the optimal alignment path according to the direction of the minimum cumulative distance selected in each step.

[0102] Specifically, after completing the cumulative distance matrix, the last element of the matrix, Y[n][m], contains the minimum cumulative distance between the two sequences. Then, starting from this end point, we backtrack to the starting point (0,0), and based on the direction of the minimum cumulative distance selected at each step (up, left, or left-up), we record this series of point pairs as the final optimal alignment path.

[0103] Step S2046: Draw a heat map based on the optimal alignment path.

[0104] That is, after the calculation of the cumulative distance matrix is ​​completed, the optimal alignment path is stored in D[n+1,m+1]. In order to understand the alignment results more intuitively, a heat map of the cumulative distance matrix is ​​drawn and the alignment path is superimposed on the heat map to more intuitively show the difference and alignment of the two curves.

[0105] A heatmap is a visualization tool that uses color to represent the magnitude or density of data values. When Dynamic Time Warping is applied to compare different time series, the similarities or differences between these series can be visually displayed.

[0106] In an optional embodiment, an optimal distance matrix is ​​obtained according to the optimal alignment path, and a heat map is drawn according to the optimal distance matrix. In the heat map, the color depth of each cell represents the size of the distance between the corresponding two sequences. The darker the color, the greater the distance, that is, the lower the similarity; and vice versa.

[0107] In another optional implementation, the optimal alignment path may be presented in the form of an image, which helps to intuitively understand how DTW achieves sequence matching by adjusting the time axis.

[0108] That is to say, a heat map can be drawn according to the optimal distance matrix, or a heat map can be drawn according to the optimal alignment path. It should be noted that drawing a heat map according to the optimal distance matrix is ​​a preferred solution. This is because the heat map drawn according to the optimal distance matrix shows the cumulative distances of all possible paths in the entire alignment process, reflects the global similarity structure between sequences, and can intuitively display the alignment difficulty of different regions (such as high-cost areas). In addition, the heat map drawn according to the optimal distance matrix can reveal multiple potential alignment paths (such as local minimum areas), helping to analyze whether there are multiple reasonable alignment methods.

[0109] Step S205: Input the image into the trained convolutional neural network model to obtain a first estimated value of the current battery health state.

[0110] In an optional embodiment, the convolutional neural network model includes a convolution layer, a pooling layer, a flattening layer, and a fully connected layer. The convolution layer is used to extract features from the image to obtain a feature map; the pooling layer is used to pool the feature map; the flattening layer is used to convert the pooled feature map into a one-dimensional vector; and the fully connected layer is used to obtain a first estimate based on the one-dimensional vector.

[0111] The convolutional layer automatically learns and extracts useful features from the input image, which is particularly important for battery health monitoring. Through convolution, the model captures complex patterns and nonlinear features in image data such as battery charge / discharge curves and temperature profiles, which are crucial for assessing battery health. The pooling layer reduces the spatial size of the feature map through downsampling, reducing the complexity of subsequent computations and the number of model parameters. It also helps preserve the most important features and makes the model more robust to small changes in the input data, which is very useful for handling data fluctuations that may occur during battery operation. The flattening layer converts the multi-dimensional feature map after pooling into a one-dimensional vector, making the data structure simpler and more straightforward for further processing by the fully connected layer. This step effectively bridges the gap between the feature representation generated by the convolutional operation and traditional machine learning methods. The fully connected layer uses the one-dimensional feature vector extracted from the image to predict a first estimate of the battery value, such as the remaining service life or the current health score. The rational design of the entire network architecture, especially the feature extraction and dimensionality reduction steps mentioned above, makes the final estimate more accurate and reliable. This CNN-based model can adapt to different battery types and specifications and can be applied to different scenarios by simply adjusting model parameters or retraining. In addition, the model can also process input data in multiple formats, such as images of different resolutions or time series data, improving the model's versatility and flexibility.

[0112] Specifically, after feature extraction in the convolutional layer, the output feature map is passed to the pooling layer for feature selection and information filtering. The pooling layer contains a pre-defined pooling function, which replaces the result of a single point in the feature map with the feature map statistics of its adjacent area. The activation function is ReLU (Rectified Linear Unit). After the convolution and pooling layers, the matrix shape is flattened into a single vector containing all the information required for prediction.

[0113] Specifically, in the convolutional neural network model, there will be an error between the estimated value and the true value. The convolutional neural network model corrects the estimated value of SOH based on this error.

[0114] The evaluation criteria between the estimated value and the true value can be the root mean square error RMSE, the mean absolute error MAE, or the mean absolute percentage error MAPE.

[0115] Specifically,

[0116]

[0117] Among them, in the calculation formulas of RMSE, MAE and MAPE above, y ′is the estimated value, y is the true value, and N is the number of estimated values / true values.

[0118] In an optional embodiment, the convolutional neural network model further includes a shielding layer, which is provided before the convolutional layer and is used to preprocess the image to shield alignment paths in the image that are away from the diagonal lines. This can increase the accuracy of the convolutional neural network model in feature extraction.

[0119] This embodiment provides a battery state of health (SOH) estimation method that uses the DTW algorithm to perform a differential analysis between the battery's current capacity increment curve and initial capacity increment curve, converting these differences into a visual image. A convolutional neural network is then used to extract features from the image, enabling high-precision estimation of the battery's state of health (SOH). This method converts time series data into image data, enabling the convolutional neural network model to efficiently extract the nonlinear characteristics of the battery's state of health, significantly improving the accuracy of SOH estimation. Furthermore, by providing a shielding layer within the convolutional neural network model, the accuracy of the convolutional neural network model during feature extraction can be further improved.

[0120] In this embodiment, a battery health status estimation method is provided, which can be used in the above-mentioned computer device, such as a cloud platform. Figure 7 is a flowchart of a battery health status estimation method according to an embodiment of the present invention. Figure 8 FIG. 1 is a flow chart of a method for estimating the SOH of a power battery based on a battery twin model and CNN according to an embodiment of the present invention. Figure 7 and Figure 8 As shown, the process includes the following steps:

[0121] Step S701: Obtain battery management parameters during the battery charging phase.

[0122] Step S702: Determine the current capacity increment curve according to the battery management parameters.

[0123] Step S703: Obtaining the initial capacity increment curve of the battery.

[0124] Step S704: Calculate the difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm, and convert the difference into an image.

[0125] Step S705: Input the image into the trained convolutional neural network model to obtain a first estimated value of the current battery health state.

[0126] Step S706: Input the battery management parameters into the preset twin model to obtain a second estimated value of the current battery health status.

[0127] Figure 9: is a schematic diagram of an example of a battery twin model according to an embodiment of the present invention, as shown in FIG. Figure 9 As shown in Figure 1, the battery twin model includes all the basic components of a lithium-ion battery, including positive and negative electrodes, positive and negative current collectors, electrode liquid, and separator.

[0128] Step S707: using the second estimated value to correct the first estimated value to obtain a third estimated value of the current battery health state.

[0129] After obtaining the first estimated value of the current battery health status (based on convolutional neural network image recognition), this embodiment further inputs the battery management parameters into the preset electrochemical twin model to obtain the second estimated value, and then corrects the two estimated results through a weighted fusion strategy to obtain the final third estimated value. This design that combines the model-based method (Model-based) with the data-driven method (Data-driven) effectively makes up for the shortcomings of a single method in terms of modeling error, generalization ability or real-time performance; significantly reduces the bias and variance of the SOH estimation, and improves the overall prediction accuracy. The aging feature matrix generated by the DTW algorithm can capture the dynamic evolution of the battery differential voltage curve at different aging stages. The convolutional neural network model extracts spatial structural features from the image perspective and belongs to the macroscopic level of representation; while the electrochemical twin model simulates the internal reaction kinetics of the battery from the perspective of microscopic physical mechanisms; the two form multi-scale information complementarity, which not only retains the flexibility of data-driven, but also enhances the physical interpretability of the model, thereby improving the robustness and environmental adaptability of the entire system.

[0130] In the related art, most of the battery SOH estimation is done by using offline state data for prediction, while this embodiment estimates the SOH online. Figure 7 and Figure 8 As shown, in this embodiment, steps S701 to S707 are performed on the cloud platform. The cloud platform not only stores historical data but also enables online, dynamic estimation of SOH through continuously updated convolutional neural network models and battery twin models. Compared to traditional offline methods, this system can update the battery status in real time while the vehicle is in operation, significantly improving the real-time and practicality of SOH estimation.

[0131] This embodiment also provides a battery health status estimation device, which is used to implement the above-mentioned embodiments and preferred embodiments. Details that have already been described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0132] This embodiment provides a battery health status estimation device, such as Figure 10 As shown, including:

[0133] The first acquisition module 1001 is used to acquire battery management parameters during the battery charging and discharging phases.

[0134] The current capacity increment curve determining module 1002 is configured to determine the current capacity increment curve according to the battery management parameters.

[0135] The second acquisition module 1003 is configured to acquire an initial capacity increment curve of the battery.

[0136] The image determination module 1004 is configured to calculate the difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm, and convert the difference into an image.

[0137] The first battery health state determination module 1005 is used to input the image into the trained convolutional neural network model to obtain a first estimated value of the current battery health state.

[0138] In some optional embodiments, the battery health state estimation apparatus further includes a second battery health state determination module. After obtaining the first estimated value of the current battery health state, the second battery health state determination module is configured to input battery management parameters into a preset twin model to obtain a second estimated value of the current battery health state; and to use the second estimated value to correct the first estimated value to obtain a third estimated value of the current battery health state.

[0139] In an optional embodiment, the battery management parameters are the current-time correspondence and the voltage-time correspondence during the charge and discharge phases. The current capacity increment curve determination module 1002 is specifically configured to: determine the capacity charged into the battery at each time point during the charge and discharge phases based on the current-time correspondence to obtain a capacity-time correspondence; determine a voltage-capacity curve based on the voltage-time correspondence and the capacity-time correspondence; and perform a differential process on the voltage-capacity curve to obtain the current capacity increment curve.

[0140] In an optional embodiment, the image determination module 1004 is specifically used to: use the current capacity increment curve as the query sequence Q of the dynamic time warping algorithm, and use the initial capacity increment curve as the template sequence C of the dynamic time warping algorithm; construct a distance matrix D, where the current element D[i][j] in the distance matrix D represents Q i and C jThe distance between them; create a cumulative distance matrix Y, where the current element Y[i][j] in the cumulative distance matrix Y is used to store the minimum cumulative distance from the starting point (0, 0) to the current point (i, j); use dynamic programming principles and path constraints to fill the cumulative distance matrix Y, where the path constraints include boundary conditions and window constraints, and the window constraints are used to pass through the range of the window size limit j in the process of filling the cumulative distance matrix Y; after the cumulative distance matrix Y is filled, trace back from the end point to the starting point, and obtain the optimal alignment path according to the direction of the minimum cumulative distance selected at each step; determine the optimal distance matrix based on the optimal alignment path, and draw an image based on the optimal distance matrix.

[0141] In an optional embodiment, the convolutional neural network model includes a convolution layer, a pooling layer, a flattening layer and a fully connected layer; the convolution layer is used to extract features of the image to obtain a feature map; the pooling layer is used to pool the feature map; the flattening layer is used to convert the pooled feature map into a one-dimensional vector; the fully connected layer is used to obtain a first estimated value based on the one-dimensional vector.

[0142] In an optional embodiment, the method for determining the window size in the window constraint includes: determining the data volume according to the current capacity increment curve and the initial capacity increment curve; determining the window size according to the data volume, wherein the window size is positively correlated with the data volume.

[0143] In an optional embodiment, the convolutional neural network model further includes a shielding layer, which is disposed before the convolutional layer and is used to preprocess the image to shield alignment paths in the image that are away from the diagonal line.

[0144] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0145] The battery health status estimation device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0146] The embodiment of the present invention also provides a computer device having the above Figure 10 The battery health status estimation device shown.

[0147] See also Figure 11 , Figure 11 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 11As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 11 A processor 10 is taken as an example.

[0148] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0149] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0150] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0151] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0152] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 11 The bus connection is taken as an example.

[0153] The input device 30 can receive input digital or character information and generate key signal input related to user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touch pad, an indicator stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 can include a display device, an auxiliary lighting device (e.g., an LED), and a tactile feedback device (e.g., a vibration motor). The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display, and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0154] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0155] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0156] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for estimating a battery health state, characterized in that: include: Obtain battery management parameters during the battery charging and discharging stages; determining a current capacity increment curve according to the battery management parameters; Obtaining an initial capacity increment curve of the battery; calculating a difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm, and converting the difference into an image; The image is input into a trained convolutional neural network model to obtain a first estimated value of the current battery health state.

2. The method according to claim 1, characterized in that After obtaining the first estimated value of the current battery health state, the following steps are also included: Inputting the battery management parameters into a preset twin model to obtain a second estimated value of the current battery health state; The first estimated value is corrected using the second estimated value to obtain a third estimated value of the current battery health state.

3. The method according to claim 1, characterized in that The battery management parameters are the current-time correspondence and the voltage-time correspondence during the charge and discharge phases, and determining the current capacity increment curve according to the battery management parameters includes: Obtaining the capacity of the battery at each time point during the charge and discharge phase according to the current-time correspondence relationship, and obtaining a capacity-time correspondence relationship; Obtaining a voltage-capacity curve according to the voltage-time correspondence relationship and the capacity-time correspondence relationship; The voltage-capacity is differentiated to obtain the current capacity increment curve.

4. The method according to claim 1, wherein The calculating the difference between the current capacity increment curve and the initial capacity increment curve by using a dynamic time warping algorithm and converting the difference into an image includes: Using the current capacity increment curve as the query sequence Q of the dynamic time warping algorithm and using the initial capacity increment curve as the template sequence C of the dynamic time warping algorithm; Construct a distance matrix D, where the current element D[i][j] in the distance matrix D represents Q i and C j the distance between them; Create a cumulative distance matrix Y, wherein the current element Y[i][j] in the cumulative distance matrix Y is used to store the minimum cumulative distance from the starting point (0, 0) to the current point (i, j); Filling the cumulative distance matrix Y using dynamic programming principles and path constraints, wherein the path constraints include boundary conditions and window constraints, wherein the window constraints are used to pass through a range of window size limits j during the process of filling the cumulative distance matrix Y; After the cumulative distance matrix Y is filled, backtrack from the end point to the starting point, and obtain the optimal alignment path according to the direction of the minimum cumulative distance selected at each step; The image is drawn according to the optimal alignment path.

5. The method according to claim 4, characterized in that Drawing the image according to the optimal alignment path includes: Obtaining an optimal distance matrix according to the optimal alignment path, and drawing a heat map according to the optimal distance matrix; or; A heat map is drawn according to the optimal alignment path.

6. The method according to claim 4, characterized in that The method for determining the window size in the window constraint includes: determining the data volume according to the current capacity increment curve and the initial capacity increment curve; The window size is determined according to the data amount, wherein the window size is positively correlated with the data amount.

7. The method according to claim 1, characterized in that The convolutional neural network model includes a convolution layer, a pooling layer, a flattening layer and a fully connected layer; The convolution layer is used to extract features from the image to obtain a feature map; The pooling layer is used to perform pooling processing on the feature map; The flattening layer is used to convert the feature map after pooling into a one-dimensional vector; The fully connected layer is used to obtain the first estimated value according to the one-dimensional vector.

8. The method according to claim 7, characterized in that The convolutional neural network model also includes a shielding layer, which is arranged before the convolution layer and is used to preprocess the image to shield the alignment path in the image away from the diagonal line.

9. A battery health status estimation device, characterized in that: include: A first acquisition module is used to obtain battery management parameters during the battery charging and discharging stages; a current capacity difference increment curve determining module, configured to determine a current capacity increment curve according to the battery management parameters; A second acquisition module is used to obtain an initial capacity increment curve of the battery; an image determination module, configured to calculate a difference between the current capacity increment curve and the initial capacity increment curve using a dynamic time warping algorithm, and convert the difference into an image; The first battery health state determination module is used to input the image into a trained convolutional neural network model to obtain a first estimated value of the current battery health state.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the battery health status estimation method according to any one of claims 1 to 8 by executing the computer instructions.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the battery health state estimation method according to any one of claims 1 to 8.

12. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the battery health state estimation method according to any one of claims 1 to 8.

Images