A method for estimating SOH of a battery pack

Abstract

This application proposes a battery pack SOH estimation method, including: S1: obtaining the capacity label value required to estimate the battery pack's SOH based on the historical actual operation data of the battery pack, and constructing a sliding window input data matrix related to the SOH; S2: based on the sliding window input data matrix, automatically extracting health feature maps related to battery pack aging through a neural network; S3: based on the extracted health feature maps, modeling the battery attenuation dynamics through a PINN neural network to achieve battery SOH estimation.

Description

Technical Field

[0001] The present application relates to the field of energy storage, and in particular to a method for estimating the SOH of a battery pack. Background Art

[0002] In the field of power electronics, battery packs are widely used as energy storage elements in various scenarios, such as electric vehicles, portable electronic devices, and energy storage.

[0003] The battery pack consists of multiple single cells connected in series, parallel, or series-parallel. The battery pack also includes a Battery Management System (BMS), which manages and monitors the battery pack. For example, it collects information such as the total battery pack voltage, maximum and minimum single cell voltages, battery pack current, maximum and minimum temperatures, cumulative battery pack charge, and battery pack SOC at regular sampling intervals.

[0004] In practical applications, battery packs can age over the course of their service life. This aging is primarily manifested by capacity decay and increased internal resistance, which is related to operating conditions such as battery age, SOC, depth of discharge (DOD), charge / discharge rate, and temperature. To ensure the long-term, safe, and continuous use of battery packs, an accurate State of Health (SOH) assessment is essential.

[0005] A battery pack's SOH is defined as the ratio of its current usable capacity to its initial capacity and can be used as a measure of battery aging. For example, when a battery pack's SOH drops to 80%, it reaches the end of its first service life. Therefore, accurately estimating a battery pack's SOH is crucial.

[0006] In the existing technology, the SOH estimation of battery packs mainly includes battery model method and data-driven method. However, the model method and data-driven method have problems such as dependence on working conditions, poor interpretability, and difficulty in balancing calculation accuracy, calculation efficiency and robustness. Summary of the Invention

[0007] According to one embodiment, the present application provides a method for estimating the SOH of a battery pack, comprising:

[0008] S1: Obtain the capacity label value required to estimate the battery pack's state of health (SOH) based on the battery pack's historical actual operating data, and construct a sliding window input data matrix related to SOH. S2: Based on the sliding window input data matrix, a neural network is used to automatically extract health feature maps related to battery pack aging.

[0009] S3: Based on the extracted health feature map, the battery degradation dynamics are modeled through the PINN neural network to estimate the battery SOH.

[0010] Furthermore, step S1 also includes: S11: obtaining historical actual operation data of the battery pack; S12: extracting a charging data set of the battery pack from the historical actual operation data, and dividing the charging data set into multiple charging segments; S13: calculating the capacity value of each charging segment, the capacity values ​​of multiple charging segments form a capacity value sequence, and obtaining the capacity label value of each charging segment; S14: calculating the capacity increment value at each moment of each charging segment, and obtaining the capacity increment sequence of each charging segment; S15: normalizing the m variables and capacity increment values ​​Q(t) related to the SOH value in each charging segment; S16: using a sliding window mechanism to divide the m variables and capacity increment values ​​Q(t) related to the SOH value in each charging segment after normalization, and obtaining multiple n×m sliding window input data matrices under each charging segment, where n is the window size, m is the number of columns, representing the number of variable channels, and n and m are both natural numbers greater than or equal to 1.

[0011] Furthermore, the historical actual operation data in step S11 includes a corresponding database of sampling time and battery pack total voltage, maximum single cell voltage, minimum single cell voltage, battery pack current, maximum temperature, minimum temperature, battery pack cumulative charging capacity and battery pack SOC.

[0012] Furthermore, step S12 includes: determining the state of each frame of data based on the battery pack current and battery pack SOC changes in the historical actual operation data, and extracting the battery pack charging data set from the historical actual operation data; for the charging data set, dividing the charging segments according to the continuity of the changes in the parameters characterizing the battery pack charge.

[0013] Furthermore, step S12 further includes: retaining the charging segments in which the lowest SOC is less than a first threshold and the highest SOC is greater than a second threshold.

[0014] Furthermore, in step S13, according to the formula Calculate the capacity value C of each charging segment c , where I(t) is the charging current of the current charging segment, SOC1 is the selected first SOC value, SOC2 is the selected second SOC value, and ΔSOC is the difference between the second SOC value and the first SOC value.

[0015] Furthermore, step S13 further includes: performing a linear fit on the capacity value sequence with respect to the accumulated charge quantity of the battery pack based on a local weighted linear regression algorithm to obtain a denoised capacity value sequence.

[0016] Furthermore, according to the formula The capacity increment value at each moment of each charging segment is obtained, where Δt is the sampling interval.

[0017] Furthermore, step S15 is: according to the formula Normalize the m variables related to the SOH value and the capacity increment value Q(t) in each charging segment, where X represents the data of the current normalized object dimension, X max and X min Indicates the maximum and minimum values ​​of the current normalized object in the battery pack's charge dataset.

[0018] Furthermore, S2 includes: S21: inputting each sliding window input data matrix into the global feature extraction network, and outputting the global features optimized by channel attention and spatial attention; S22: performing spatial graph feature extraction on the global features through the graph feature extraction network to obtain a health feature map related to battery pack aging.

[0019] Furthermore, step S22 is: performing spatial graph feature extraction on the global features through the GAT+GCN model to obtain a health feature graph related to battery pack aging.

[0020] Furthermore, S3 includes: S31: inputting the health feature map into a multi-layer fully connected neural network and outputting an SOH estimation value; S32: constructing a PINN neural network so that the partial derivative of the function defined by the multi-layer fully connected neural network with respect to time is equal to the output of the PINN neural network; S33: training the global feature extraction network, the graph feature extraction network, the multi-layer fully connected neural network, and the PINN neural network.

[0021] Furthermore, the input of the PINN neural network is the SOH estimate SOH and the partial derivative of the SOH estimate with respect to time SOH t , the partial derivative SOH of the SOH estimate with respect to the health characteristic map G , health feature map G and time t, the output of the PINN neural network is the decay rate of SOH. The PINN neural network makes the partial derivative of the function defined by the multi-layer fully connected neural network with respect to time equal to the decay rate of SOH.

[0022] The features and technical advantages of the present disclosure have been summarized quite broadly above so that the detailed description disclosed below may be better understood. Additional features and advantages of the present disclosure will be described below, which form the subject matter of the claims of the present disclosure. It will be understood by those skilled in the art that the concepts and specific embodiments disclosed herein may be readily used as a basis for modifying or designing other structures or processes for achieving the same purposes of the present disclosure. It will also be appreciated by those skilled in the art that such equivalent structures do not depart from the spirit and scope of the present disclosure as set forth in the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] For a more complete understanding of the present disclosure and its advantages, reference is now made to the following description in conjunction with the accompanying drawings:

[0024] in:

[0025] Figure 1 A schematic flow chart of a method for estimating the SOH of a battery pack according to an embodiment of the present application is shown;

[0026] Figure 2 A schematic diagram of the data processing process of an embodiment of the present application is shown;

[0027] Figure 3 A schematic diagram of a process for dividing charging segments according to an embodiment of the present application is shown;

[0028] Figure 4 A schematic diagram of the SOH estimation principle of a battery pack according to an embodiment of the present application is shown.

[0029] Corresponding numerals and symbols in the different figures generally refer to corresponding parts unless otherwise indicated. These figures are drawn to clearly illustrate the relevant aspects of the various embodiments and are not necessarily drawn to scale. DETAILED DESCRIPTION

[0030] The following is a clear and complete description of the technical solutions in this application in conjunction with the accompanying drawings. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0031] One embodiment of the present application is to provide a method for estimating the SOH of a battery pack. Figure 1 The flowchart of the battery pack SOH estimation method according to one embodiment of the present application is shown. The battery pack SOH estimation method proposed in the present application includes:

[0032] S1: Obtain the capacity label value required to estimate the SOH of the battery pack based on the historical actual operation data of the battery pack, and construct a sliding window input data matrix related to the SOH;

[0033] S2: Based on the sliding window input data matrix, the neural network automatically extracts the health feature map related to battery pack aging;

[0034] S3: Based on the extracted health feature map, the battery degradation dynamics are modeled through the PINN neural network to estimate the battery SOH.

[0035] See also Figure 2 The data processing stage flow diagram of an embodiment of the present application is shown in FIG. Specifically, step S1 includes:

[0036] S11: Acquire historical actual operating data of the battery pack; S12: Extract a charging data set of the battery pack from the historical actual operating data and divide the charging data set into multiple charging segments; S13: Calculate the capacity value of each charging segment, form a capacity value sequence from the capacity values ​​of the multiple charging segments, and obtain a capacity label value for each charging segment; S14: Calculate the capacity increment value at each moment of each charging segment to obtain a capacity increment sequence for each charging segment; S15: Normalize m variables related to the SOH value and the capacity increment value Q(t) within each charging segment;

[0037] S16: Use a sliding window mechanism to divide the m variables related to the SOH value and the capacity increment value Q(t) in each charging segment after normalization, and obtain multiple n×m sliding window input data matrices under each charging segment, where n is the window size and m is the number of columns, representing the number of variable channels. Both n and m are natural numbers greater than or equal to 1.

[0038] For step S11, specifically, during the actual operation of the battery pack, data including but not limited to the total voltage of the battery pack, the highest single cell voltage, the lowest single cell voltage, the battery pack current, the highest temperature, the lowest temperature, the cumulative charge capacity of the battery pack, and the battery pack SOC, etc., will be obtained at a certain sampling interval (such as 10s, but not limited to this, and can be determined according to different application scenarios). Therefore, the historical actual operation data in step S1 is the actual data of the battery pack in the real world, which is formed during the historical operation process. For example, the historical actual operation data includes a corresponding database of sampling time and the total voltage of the battery pack, the highest single cell voltage, the lowest single cell voltage, the battery pack current, the highest temperature, the lowest temperature, the cumulative charge capacity of the battery pack, and the battery pack SOC. The historical actual operation data includes the cumulative charge capacity of the battery pack, as well as the highest single cell voltage, the lowest single cell voltage, the highest temperature, and the lowest temperature that reflect the inconsistency of the cells, which can improve the SOH estimation accuracy of the battery pack.

[0039] Furthermore, in step S11, when abnormal or missing values ​​appear in the historical actual operation data, the historical actual operation data is corrected. Specifically, when the missing time frames are small, the SOC is monotonic or constant, or the current is constant, linear interpolation is used to fill in the missing values ​​to correct the historical actual operation data. When the missing time frames are large (e.g., the data is missing for more than 50 seconds) and the missing segments are in the middle of the charging process, the charging cycle data is discarded to correct the historical actual operation data.

[0040] For step S12, specifically, it includes: determining the state of each frame of data based on the battery pack current and battery pack SOC changes in the historical actual operation data, extracting the battery pack charging data set from the historical actual operation data; then, for the charging data set, dividing different charging cycles according to the continuity of the change of the parameter (such as SOC) that characterizes the battery pack charge, and a charging cycle can be called a charging segment. If the battery pack SOC continuously increases from time t1 to time t2 and then decreases, the time from t1 to time t2 is considered to be a charging segment. For details, please refer to Figure 3 The flowchart of dividing the charging segments according to an embodiment of the present application is shown in FIG. Figure 3 As shown, the charging data set is divided into multiple charging segments, including: S121: receiving a charging data set; S122: defining start=1, t=1; S123: judging whether the SOCt at the previous moment is greater than the SOCt+1 at the current moment, if so, proceeding to S124, if not, proceeding to S126; S124: the charging data from the start moment to the t moment in the charging data set is divided into a charging segment; S125: updating start=t+1, proceeding to S127; S126: updating t=t+1, proceeding to step S127; S127: judging whether t+1 is greater than the length of the charging data set, if so, ending, if not, proceeding to step S123. In this way, if the SOCt of the previous moment is greater than the SOCt+1 of the current moment, for example, the SOC suddenly drops from 90% to 10%, then the current charging cycle is over, and the charging data from the start time (start) to the current moment t can be divided into a charging segment, and the start time (start) is updated to t+1, and the charging data of subsequent moments are continued to be judged; if the SOCt of the previous moment is less than the SOCt+1 of the current moment, for example, the SOC increases and drops from 88% to 90%, then the current charging cycle is still in progress, and the time t is updated to t+1, and the charging data of subsequent moments are continued to be judged until the last moment of the charging data set, so that the charging data set is divided into multiple charging segments.

[0041] In actual operation, step S12 further includes screening charging segments. To cover as many charging segments as possible while ensuring that each charging segment includes the battery pack voltage non-plateau region, specifically, charging segments with a minimum SOC less than a first threshold (e.g., 40%) and a maximum SOC greater than a second threshold (e.g., 70%) are retained. Charging segments that do not meet this condition are considered inappropriate and are removed.

[0042] For step S13, specifically, according to the formula Calculate the capacity value C of each charging segment c , where I(t) is the charging current of the current charging segment, SOC1 is the selected first SOC value, and SOC2 is the selected second SOC value. It is usually desired that the range from the first SOC value to the second SOC value includes the voltage platform area of ​​the battery pack and also includes a portion of the voltage non-platform area. For example, in one embodiment, the first SOC value can be 40%, the second SOC value can be 70%, and ΔSOC is the difference between the second SOC value and the first SOC value.

[0043] In actual operation, step S13 further includes: noise processing of the capacity value sequence. Specifically, a linear fit of the capacity value sequence with respect to the battery pack's cumulative charge capacity is performed using a local weighted linear regression algorithm to obtain a denoised capacity value sequence. This is because the cumulative number of cycles cannot accurately reflect the battery pack's usage level.

[0044] In practical applications, the denoised capacity value sequence can be used as a label value for subsequent SOC estimation model training. Alternatively, the capacity value sequence obtained in step S13 can be directly used as a label value for subsequent SOC estimation model training.

[0045] For step S14, specifically, according to the formula The capacity increment value at each moment of each charging segment is obtained, where Δt is the sampling interval. At the start time t=0, Q(t)=0. The capacity increment can represent the characteristics of the battery pack.

[0046] For step S15, specifically, according to the formula For each charging segment, there are m variables related to the SOH value, such as the charging current I(t), the total voltage of the battery pack V total (t), the highest single voltage V max (t), minimum single cell voltage V min (t), maximum temperature T max (t), minimum temperature T min (t), and the capacity increment value Q(t) are normalized, where X represents the data of the current normalized object dimension, X max and X minIndicates the maximum and minimum values ​​of the current normalized object in the battery pack charging data set, such as normalizing the charging current I(t), X max Indicates the maximum value of the charging current in the battery pack’s charging data set, X min Indicates the minimum value of the charging current in the battery pack's charging data set.

[0047] For step S16, specifically, in order to enhance the highly relevant part of the data reading information and better capture the time correlation of the data, a sliding window mechanism is introduced. The seven variables related to the SOH value are the charging current I(t), the total voltage of the battery pack V total (t), the highest single voltage V max (t), minimum single cell voltage V min (t), maximum temperature T max (t), minimum temperature T min (t) as an example, in an embodiment of practical application, the window size n can take an initial value of 500, then a sliding window input data matrix X input It can be expressed as X input =[V total (t),V max (t),V min (t),I(t),T max (t),T min (t),Q(t)] 500×7 The number of windows can be initialized to 3, meaning that three 500×7 sliding windows are used as the input data matrix in each charging segment. 500 is the window size, i.e., the number of rows in the sliding window input data matrix, and 7 is the number of variable channels, with one column representing each variable channel. In one embodiment of a subsequent practical application, the parameters of the trainable validation pair are optimized using a grid search within the window size range of {300, 400, 500} and the number of windows range of {3, 5, 7}, using the validation set loss as the evaluation metric.

[0048] Thus, for one embodiment of the present application, the sliding window input data matrix in step S1 is the sliding window input data matrix formed in step S16. The sliding window input data matrix of the present application takes 7 variable channels as an example. Of course, the number of variables can be increased or decreased, and different variables can be used according to different needs or application scenarios.

[0049] Step S2 includes: S21: inputting each sliding window input data matrix into the global feature extraction network, and outputting the global features optimized by channel attention and spatial attention; S22: performing spatial graph feature extraction on the global features through the graph feature extraction network to obtain a health feature map related to battery pack aging.

[0050] In one embodiment, the global feature extraction network in step S21 is a CNN-LSTM+CBAM hybrid module. Figure 4 The schematic diagram of the SOH estimation principle of a battery pack according to an embodiment of the present application is shown. In practical applications, the importance of each parameter in the sliding window input data matrix to the final SOH is different. Therefore, each sliding window input data matrix X input The input is fed into a CNN-LSTM+CBAM hybrid module, a deep learning hybrid model that combines a convolutional neural network (CNN), a long short-term memory network (LSTM), and a convolutional block attention module (CBAM). Its task is to apply weights. Through deep learning, this CNN-LSTM+CBAM hybrid module obtains global features corresponding to each sliding window input data matrix. These global features are weight matrices with the same dimensions (number of rows and columns) as the sliding window input data matrix, with values ​​between 0 and 1. This enhances features with strong correlation to the state of interest (SOH) and variable channels within the sliding window input data matrix, while simultaneously weakening features with weak SOH correlation and variable channels within the sliding window input data matrix. This allows the CNN-LSTM+CBAM hybrid module to focus attention on features with strong SOH correlation and suppress features with weak SOH correlation.

[0051] For step S22, Figure 4 As shown in the figure, the GAT+GCN model is used to extract spatial graph features from global features to obtain a health feature map related to battery pack aging. The GAT+GCN model combines the Graph Attention Network (GAT) and the Graph Convolutional Network (GCN). Specifically, the Graph Attention Network (GAT) uses each variable channel of the global feature as a graph node. Each node is connected to its k nearest neighbor nodes to form an edge. The adjacency matrix A is constructed based on the node and edge information. The calculation formula is:

[0052]

[0053] Among them, node A ij For node b i With node b j The connection relationship between them, K(b i ) is node b i The nodes in the k nearest neighbors of . And through the attention mechanism of GAT, the updated node b′ is obtained i, the formula is:

[0054] b′ i =α ii θb i +∑ j∈N(i) α ij θb j (2)

[0055] Among them, θ is the global feature (that is, the weight matrix), b i and b j is the node feature in the global feature, b′ i is the updated node feature, α ij It represents the attention weight coefficient of each node i to each of its neighbor nodes j. When j is equal to i, it represents the attention weight coefficient of node i to itself. The formula is:

[0056]

[0057] Here, LeakyReLu(·) is a nonlinear activation function, a is a weight vector used to represent attention, || is a connection operation, and N(i) is the neighbors of node i. GAT introduces an attention mechanism, allowing the model to automatically learn the importance weights of different neighboring nodes to the current node, thereby more effectively capturing the complex relationships between nodes. The output features of GAT are further input into the Graph Convolutional Network (GCN) to obtain the final health feature map related to battery pack aging. The health feature map can be expressed as G = {V, E, A}, where V is the node set, E is the edge set connecting any two nodes, and A is the adjacency matrix.

[0058] In this way, data processing, cleaning and filtering are completed, and the health feature map related to battery pack aging is extracted using the neural network layer.

[0059] Step S3 includes: S31: inputting the health feature map into a multi-layer fully connected neural network and outputting an SOH estimation value; S32: constructing a PINN neural network so that the partial derivative of the function defined by the multi-layer fully connected neural network with respect to time is equal to the output of the PINN neural network; S33: training the global feature extraction network, the graph feature extraction network, the multi-layer fully connected neural network, and the PINN neural network.

[0060] For step S31, specifically, based on the health feature map extracted and fused by CNN-LSTM+CBAM+GAT+GCN, a multi-layer fully connected neural network (FCNN or FC) is used to learn and extract the nonlinear relationship between the health feature map (i.e., aging features) and SOH, and output the SOH estimation value.

[0061] A multi-layer fully connected neural network (FC) can be defined as a network with learnable parameters Φ Used to characterize the battery aging model f(t,G), where the battery aging trajectory is modeled as:

[0062] SOH=f(t,G) (4)

[0063] Wherein, t represents time, and G represents the health feature map extracted above.

[0064] Physics-Informed Neural Networks (PINN) is a new type of machine learning model that combines deep learning and physical laws. It embeds physical laws (usually expressed in the form of partial differential equations or ordinary differential equations) into the loss function of the neural network, so that the network not only relies on data during training, but also satisfies the laws of physics. For step S32, specifically, the input of the PINN neural network is the SOH estimate SOH, the partial derivative of the SOH estimate with respect to time SOH t , the partial derivative SOH of the SOH estimate with respect to the health characteristic map G , health feature map G and time t, the output of the PINN neural network is the decay rate of SOH. The PINN neural network makes the partial derivative of the function defined by the multi-layer fully connected neural network (FC) with respect to time equal to the decay rate of SOH. Specifically, the decay rate of SOH is expressed as is the battery degradation dynamic equation of the PINN neural network, Θ is the learnable parameter of the PINN neural network, and the battery aging mechanism is learned from the given battery degradation aging data set to explicitly construct the nonlinear relationship between battery SOH and many influencing factors. The partial derivative of the function defined by the multi-layer fully connected neural network (FC) with respect to time is Characterize the battery pack degradation trajectory, that is,

[0065] In this way, the decay dynamic characteristics of the battery pack are learned through the established PINN neural network, and the learning results are made closer to the actual situation of the battery pack, thereby achieving the purpose of training the PINN neural network.

[0066] For step S33, specifically, according to the formula:

[0067]

[0068] Calculate the SOH tag value of the battery pack at each charging segment, where Q k represents the capacity label value under the kth charging segment, and Q0 represents the rated capacity.

[0069] For characterizing battery degradation trajectory Φ and aging decay dynamics in The Θ in θ is trained by minimizing the mean squared error loss, and the loss function is:

[0070]

[0071] in, The optimization process of the PINN neural network needs to be followed The concrete PDE loss formula is:

[0072]

[0073] Wherein, the subscript i represents the i-th charging segment, G i Represents the health feature map under the i-th charging segment.

[0074] is the data fitting loss, and the formula is:

[0075]

[0076] Wherein, the subscript i represents the i-th charging segment, N represents the number of charging segments, Indicates the label value under the i-th charging segment, SOH i represents the estimated value under the i-th charging segment.

[0077] is the monotonicity loss of the SOH trajectory, and the formula is:

[0078]

[0079] Among them, the monotonicity loss Based on the physical characteristics of battery degradation, the SOH of the next cycle should be less than or equal to the SOH of the previous cycle (unless capacity regeneration occurs), where represents the label value in the k+1th charging segment, Indicates the label value of the kth charging segment.

[0080] In this way, the SOH value output by the neural network is compared with the label value. If the loss is too large, the parameters Φ and Θ in the neural network will be gradient updated so that the SOH value output by the neural network in each segment is closer to the label value.

[0081] The SOH estimation method provided in this application is based on the real-world battery pack operating data, and the basic data is cleaned and the capacity label value is calculated to obtain the basic input data of the neural network model. In order to make the neural network model have better adaptability to working conditions, based on the end-to-end idea, the CNN-LSTM and the Convolutional Block Attention Module (CBAM) are used to initially extract features, and then a multi-layer graph neural network (GNN) is constructed to capture the complex nodes and edge relationships in the graph, and autonomously learn to extract aging-related features. Finally, the pre-order network is integrated with the physical information neural network training model to realize the battery pack SOH estimation. It can be seen that the method provided in this application does not rely on the constant current charging working condition requirements, and at the same time fully combines the advantages of the battery pack physical information model and the data-driven method to estimate the battery pack SOH, solving the imbalance problem between the traditional model method and the data-driven algorithm in terms of working condition adaptability, computational efficiency and computational accuracy.

[0082] The above-mentioned battery pack can be a lithium iron phosphate battery, a ternary lithium battery, etc., and this application does not limit this.

[0083] The above-mentioned battery pack can be used in any occasion, such as new energy vehicles and photovoltaic energy storage, wind power generation energy storage and other actual use scenarios of lithium-ion batteries.

[0084] In practical applications, the battery pack SOH estimation method provided in this application is executed by a cloud computing platform. Cloud computing platforms have relatively high memory and computational efficiency, making them suitable for model training on large amounts of data. In practical applications, the battery management system collects and obtains the battery pack's operating data during operation. Therefore, the cloud computing platform obtains the aforementioned historical actual operating data from the battery management system. In practical applications, the model trained by the cloud computing platform can also be transmitted to the battery management system, which then obtains the battery pack's SOH estimate.

[0085] Although the embodiments of the present disclosure and its advantages have been described in detail, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the disclosure as defined by the appended claims.

[0086] Furthermore, the scope of the present application is not intended to be limited to the particular embodiments of the processes, machines, manufacture, compositions of matter, means, methods, and steps described in the specification. As will be readily understood by one of ordinary skill in the art from the disclosure herein, processes, machines, manufacture, compositions of matter, means, methods, or steps that perform substantially the same functions now exist or will later be developed or that achieve substantially the same results as the corresponding embodiments described herein may be utilized in accordance with the present disclosure. Accordingly, the appended claims are intended to include within their scope such processes, machines, manufacture, compositions of matter, means, methods, or steps.

Claims

1. A method for estimating the SOH of a battery pack, characterized in that: include: S1: Obtain the capacity label value required to estimate the SOH of the battery pack based on the historical actual operation data of the battery pack, and construct a sliding window input data matrix related to the SOH; S2: Based on the sliding window input data matrix, the neural network automatically extracts the health feature map related to battery pack aging, including: S21: Input each sliding window input data matrix into the global feature extraction network, and output the global features optimized by channel attention and spatial attention; S22: Perform spatial graph feature extraction on the global features through the graph feature extraction network to obtain a health feature graph related to battery pack aging; S3: Based on the extracted health feature map, the battery degradation dynamics are modeled through the PINN neural network to estimate the battery SOH, including: S31: Input the health feature map into a multi-layer fully connected neural network and output an SOH estimation value; S32: Construct a PINN neural network so that the partial derivative of the function defined by the multi-layer fully connected neural network with respect to time is equal to the output of the PINN neural network; S33: Train the global feature extraction network, the graph feature extraction network, the multi-layer fully connected neural network, and the PINN neural network, wherein the input of the PINN neural network is the SOH estimate SOH and the partial derivative of the SOH estimate with respect to time SOH t , the partial derivative SOH of the SOH estimate with respect to the health characteristic map G , health feature map G and time t, the output of the PINN neural network is the decay rate of SOH, and the PINN neural network makes the partial derivative of the function defined by the multi-layer fully connected neural network with respect to time equal to the decay rate of SOH.

2. The battery pack SOH estimation method according to claim 1, characterized in that: Step S1 further includes: S11: Obtain historical actual operation data of the battery pack; S12: extracting a charging data set of the battery pack from the historical actual operation data, and dividing the charging data set into a plurality of charging segments; S13: Calculate the capacity value of each charging segment, the capacity values ​​of multiple charging segments form a capacity value sequence, and obtain the capacity label value of each charging segment; S14: Calculate the capacity increment value at each moment of each charging segment to obtain the capacity increment sequence of each charging segment; S15: normalizing the m variables related to the SOH value and the capacity increment value Q(t) in each charging segment; S16: Use the sliding window mechanism to divide the m variables related to the SOH value and the capacity increment value Q(t) in each charging segment after normalization, and obtain multiple n×m sliding window input data matrices under each charging segment, where n is the window size and m is the number of columns, representing the number of variable channels. Both n and m are natural numbers greater than or equal to 1.

3. The SOH estimation method of a battery pack according to claim 2, characterized in that: The historical actual operation data in step S11 includes a corresponding database of sampling time and battery pack total voltage, maximum single cell voltage, minimum single cell voltage, battery pack current, maximum temperature, minimum temperature, battery pack cumulative charging capacity and battery pack SOC.

4. The method for estimating the SOH of a battery pack according to claim 3, wherein: Step S12 includes: Determine the state of each frame of data based on the battery pack current and battery pack SOC changes in the historical actual operation data, and extract the battery pack charging data set from the historical actual operation data; For the charging data set, multiple charging segments are divided according to the continuity of the change of the parameters representing the charge of the battery pack.

5. The battery pack SOH estimation method according to claim 4, characterized in that: Step S12 further includes: retaining the charging segment in which the lowest SOC is less than a first threshold and the highest SOC is greater than a second threshold.

6. The method for estimating the SOH of a battery pack according to claim 4, wherein: In step S13, according to the formula Calculate the capacity value C of each charging segment c , where I(t) is the charging current of the current charging segment, SOC1 is the selected first SOC value, SOC2 is the selected second SOC value, and ΔSOC is the difference between the second SOC value and the first SOC value.

7. The method for estimating the SOH of a battery pack according to claim 6, wherein: Step S13 further includes: performing a linear fit on the capacity value sequence with respect to the accumulated charge quantity of the battery pack based on a local weighted linear regression algorithm to obtain a denoised capacity value sequence.

8. The method for estimating the SOH of a battery pack according to claim 7, wherein: According to the formula The capacity increment value at each moment of each charging segment is obtained, where Δt is the sampling interval.

9. The method for estimating the SOH of a battery pack according to claim 6, wherein: Step S15 is: According to the formula Normalize the m variables related to the SOH value and the capacity increment value Q(t) in each charging segment, where X represents the data of the current normalized object dimension, X max and X min Indicates the maximum and minimum values ​​of the current normalized object in the battery pack's charge dataset.

10. The method for estimating the SOH of a battery pack according to claim 1, wherein: Step S22 is: extracting spatial graph features from global features by combining the graph attention network and the graph convolutional network model to obtain a health feature map related to battery pack aging.

Images