Method for estimating state of health of vehicle battery and method for training estimation model thereof

Abstract

The application relates to a vehicle battery health state estimation method and a training method of an estimation model thereof, wherein the training method of the vehicle battery health state estimation model comprises the following steps: obtaining battery data of a vehicle; wherein the battery data comprises first battery data and second battery data; adopting a preset clustering algorithm to cluster the first battery data, so as to divide the first battery data into multiple categories; dividing the vehicle into multiple vehicle groups according to the multiple categories of the first battery data; for each vehicle group in the multiple vehicle groups, training a battery estimation model corresponding to the vehicle group by using the second battery data corresponding to the vehicle group; obtaining the trained battery estimation model of each vehicle group; and the battery estimation model is used for estimating the vehicle battery health state in the corresponding vehicle group. Through the application, the problem that a corresponding battery estimation model is separately configured for each electric vehicle in the related art, which easily causes the problems of excessive system running load and resource waste, is solved.

Description

Technical Field

[0001] This application relates to the field of new energy vehicles, and in particular to an estimation method for the state of health of a vehicle battery and a training method for the estimation model. Background Technology

[0002] With the deterioration of the global environment and the depletion of traditional energy reserves, countries have begun to advocate for low-carbon lifestyles and establish relevant planning systems. Electric vehicles (EVs) represent a solution to the high energy consumption and pollution problems of traditional gasoline-powered vehicles. In EVs, the power battery is a crucial component, providing energy to all energy-consuming devices such as the motor, displays, and air conditioning. However, with continued use, power batteries gradually age, leading to reduced driving range and other issues that compromise the optimal driving experience. Therefore, accurately estimating the State of Health (SOH) of power batteries can not only help automakers effectively manage and improve their batteries but also provide users with guidance on battery maintenance or replacement, thus enhancing the user experience.

[0003] Current SOH estimation methods can be broadly categorized into four types: direct measurement methods, indirect analysis methods, adaptive algorithms, and data-driven methods. Direct measurement methods assess SOH by directly measuring characteristic parameters such as capacity, internal resistance, or impedance of the battery. However, these methods typically require additional equipment and are highly sensitive to environmental factors, making them more suitable for laboratory use. Indirect analysis methods calibrate SOH by measuring certain process parameters that reflect the degradation of battery capacity or internal resistance. However, these methods usually require specific stable currents or temperatures, placing high demands on external factors. Adaptive algorithms generally rely on electrochemical models or equivalent circuit models, estimating SOH by identifying model parameters; therefore, their estimation accuracy is highly dependent on the model's accuracy. Data-driven methods do not rely on precise mathematical models but instead depend on historical aging data. They use specific algorithms to learn the mapping relationship between key aging information from historical data and SOH; however, this method often requires a large amount of data.

[0004] In an existing data-driven approach, each EV's private information (vehicle ID, driving route) and public information (location, driving behavior, and battery data) in a given region are stored on an in-vehicle blockchain platform. Each EV's private information is also stored in a regional private blockchain, which then sends this information to a consortium blockchain. The consortium blockchain extracts driving behavior data (speed, acceleration, and mileage, etc.), battery status data (voltage, current, and temperature, etc.), and environmental data (temperature, humidity, and air pressure in the vehicle's location, etc.) from the private information, discarding data with 80% or more missing values ​​and interpolating data with less than 80% missing values. This data is then divided into discrete and continuous features, and similarity analysis is performed on these features, discarding features with similarity coefficients less than 0.6. The retained feature data is used to train and test an Informer model until the required accuracy is met. Subsequently, this Informer model is applied to each EV, and the Informer is recalibrated using historical data from each EV, resulting in a customized Informer model for SOH estimation for each EV.

[0005] However, the above method has the following drawbacks:

[0006] 1. The Informer model itself is very complex and has very high requirements for hardware. Configuring an Informer model for each EV in the vehicle blockchain platform would result in a large number of almost identical models. Therefore, it can easily lead to problems such as excessive equipment load and wasted resources.

[0007] 2. The data contains too many data types, and the data has not undergone in-depth feature extraction; it has only been simply divided into discrete and continuous features. This results in the subsequent SOH estimation model receiving a wide variety of low-value information, which is not conducive to the convergence of the model's loss function and thus affects the model's performance.

[0008] Currently, there is no effective solution to the problem that configuring a separate battery estimation model for each electric vehicle component in related technologies can easily lead to excessive system load and resource waste. Summary of the Invention

[0009] This invention provides a method for estimating the health status of a vehicle battery and a training method for the estimation model, in order to solve the problem in related technologies where configuring a corresponding calculation model for each electric vehicle component can easily lead to excessive equipment operating load and resource waste.

[0010] In a first aspect, this invention provides a method for training a vehicle battery health state estimation model, the method comprising:

[0011] Obtain vehicle battery data; wherein the battery data includes first battery data and second battery data;

[0012] A preset clustering algorithm is used to cluster the first battery data, dividing the first battery data into multiple categories;

[0013] The vehicles are divided into multiple vehicle groups based on the first battery data of multiple categories;

[0014] For each of the multiple vehicle groups, a battery estimation model corresponding to the vehicle group is trained using the second battery data corresponding to the vehicle group.

[0015] A trained battery estimation model is obtained for each vehicle group; wherein the battery estimation model is used to estimate the battery health status of the vehicles in the corresponding vehicle group.

[0016] In some embodiments, training the corresponding battery estimation model using the second battery data corresponding to the vehicle group includes:

[0017] Multiple autoencoder models are trained based on various second battery data corresponding to the vehicle group.

[0018] Based on the second battery data and its corresponding trained autoencoder model, a hidden vector corresponding to the second battery data is obtained;

[0019] A self-learning model is trained based on the hidden vectors;

[0020] The trained autoencoder models and the self-learning model are combined to form a battery estimation model corresponding to the vehicle group.

[0021] In some embodiments, training multiple autoencoder models based on various second battery data corresponding to the vehicle group includes:

[0022] Several data samples are determined based on various second battery data corresponding to the vehicle group; wherein each data sample includes multiple second battery data generated by the same vehicle during the same charging process;

[0023] Multiple second battery data from several data samples are respectively input into multiple autoencoder models to be trained;

[0024] Calculate the first mean squared error loss value between the input and output values ​​of the autoencoder model;

[0025] The model parameters of the autoencoder model are adjusted according to the first mean squared error loss value until the first mean squared error loss value is less than a first preset threshold, thus completing the training of the autoencoder model.

[0026] In some embodiments, obtaining the hidden vector corresponding to the second battery data based on the second battery data and its corresponding trained autoencoder model includes:

[0027] Multiple second battery data from several of the data samples are respectively input into multiple trained autoencoder models;

[0028] Obtain the hidden vector corresponding to each of the second battery data from the autoencoder model;

[0029] The self-learning model trained based on the hidden vectors includes:

[0030] Several feature samples are determined based on the hidden vectors; wherein each feature sample includes multiple hidden vectors corresponding to multiple second battery data in the same data sample;

[0031] Several of the aforementioned feature samples are input into the self-learning model to be trained;

[0032] Calculate the second mean squared error loss value between the output value of the self-learning model and the corresponding true value of the sample;

[0033] The model parameters of the self-learning model are adjusted according to the second mean squared error loss value until the second mean squared error loss value is less than the second preset threshold, thus completing the training of the self-learning model.

[0034] In some embodiments, the autoencoder model includes an encoding model and a decoding model;

[0035] The hidden vector is the output of the encoding model and the input of the decoding model;

[0036] The input to the encoding model is the input to the autoencoder model, and the output of the decoding model is the output of the autoencoder model.

[0037] In some embodiments, the first battery data includes at least one of the following:

[0038] Average charging current, average voltage at the end of charging, average state of charge, average charging temperature, charging capacity, average ambient temperature, average speed per 100 kilometers, number of accelerations per 100 kilometers, number of decelerations per 100 kilometers, number of fast charging cycles, number of slow charging cycles, cell type, and total mileage.

[0039] The second battery data includes at least one of the following:

[0040] The charging current, charging voltage, battery temperature, state of charge (SOC) value, and cumulative charge amount during slow charging, as well as the historical health status values ​​before slow charging.

[0041] Secondly, this invention provides a method for estimating the state of health of a vehicle battery, the method comprising:

[0042] Obtain battery data of the vehicle to be estimated; wherein the battery data includes first battery data and second battery data;

[0043] The vehicle group to which the vehicle to be estimated belongs is determined based on the first battery data, and the battery estimation model corresponding to the vehicle to be estimated is determined based on the vehicle group.

[0044] The battery health status of the vehicle to be estimated is determined using the battery estimation model and the second battery data.

[0045] The battery estimation model is trained using the training method for the vehicle battery health state estimation model described in the first aspect.

[0046] Thirdly, this invention provides a training apparatus for a vehicle battery health state estimation model, the apparatus comprising:

[0047] A data acquisition module is used to acquire battery data of the vehicle; wherein the battery data includes first battery data and second battery data;

[0048] The data clustering module is used to cluster the first battery data using a preset clustering algorithm, thereby dividing the first battery data into multiple categories;

[0049] A group segmentation module is used to divide the vehicle into multiple vehicle groups based on multiple types of the first battery data;

[0050] The model training module is used to train a battery estimation model corresponding to each of the multiple vehicle groups using the second battery data corresponding to the vehicle group.

[0051] The model determination module is used to obtain a trained battery estimation model for each vehicle group; wherein the battery estimation model is used to estimate the battery health status of the vehicles in the corresponding vehicle group.

[0052] Fourthly, the present invention provides a device for estimating the health status of a vehicle battery, the device comprising:

[0053] A data acquisition module is used to acquire battery data of the vehicle to be estimated; wherein the battery data includes first battery data and second battery data;

[0054] The model determination module is used to determine the vehicle group to which the vehicle to be estimated belongs based on the first battery data, and to determine the battery estimation model corresponding to the vehicle to be estimated based on the vehicle group.

[0055] A state assessment module is used to determine the battery health status of the vehicle to be estimated using the battery estimation model and the second battery data.

[0056] The battery estimation model is trained using the training device for the vehicle battery health status estimation model described in the third aspect.

[0057] Fifthly, this embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the training method for the vehicle battery health state estimation model described in the first aspect above, or the estimation method for the vehicle battery health state described in the second aspect above.

[0058] In a sixth aspect, this embodiment provides a storage medium storing a computer program that, when executed by a processor, implements the training method for the vehicle battery health state estimation model described in the first aspect above, or implements the vehicle battery health state estimation method described in the second aspect above.

[0059] Compared with related technologies, the vehicle battery health state estimation method and its estimation model training method provided in this embodiment firstly cluster the existing first battery data of all vehicles, and then divide the vehicles into multiple vehicle groups based on the clustering results. For each vehicle group, a corresponding battery estimation model is trained separately using the second battery data of that vehicle group. Then, all vehicles within a certain vehicle group can estimate their own battery health state through the corresponding battery estimation model. Compared with related technologies that configure a separate battery estimation model for each electric vehicle, this significantly reduces the number of battery estimation models, avoids generating a large number of almost homogeneous models in the battery estimation system, and thus greatly reduces the system's operating load and saves system resources. Therefore, the vehicle battery health state estimation model training method proposed in this invention solves the problem of excessive system operating load and resource waste caused by configuring a separate battery estimation model for each electric vehicle in related technologies.

[0060] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0061] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0062] Figure 1 This is a hardware structure block diagram of the terminal for the training method of the vehicle battery health state estimation model executed in this invention.

[0063] Figure 2 This is a flowchart of the training method for the vehicle battery health state estimation model in an embodiment of the present invention;

[0064] Figure 3 This is a schematic diagram of the structure of the autoencoder model in an embodiment of the present invention;

[0065] Figure 4 This is a flowchart of a method for estimating the health status of a vehicle battery in an embodiment of the present invention;

[0066] Figure 5 This is a flowchart of a method for estimating the health status of a vehicle battery in a preferred embodiment of the present invention;

[0067] Figure 6 This is a structural block diagram of the training device for the vehicle battery health state estimation model in an embodiment of the present invention.

[0068] Figure 7 This is a structural block diagram of a vehicle battery health status estimation device according to an embodiment of the present invention. Detailed Implementation

[0069] To better understand the purpose, technical solution, and advantages of this application, the application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0070] Unless otherwise defined, the technical or scientific terms used in this application shall have the general meaning as understood by one of ordinary skill in the art to which this application pertains. Words such as “a,” “an,” “an,” “the,” “the,” and “these,” used in this application, do not indicate quantitative limitation and may be singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include steps or modules (units) not listed, or may include other steps or modules (units) inherent to such processes, methods, products, or devices. The terms “connected,” “linked,” and “coupled,” used in this application, are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. The term “multiple” used in this application refers to two or more. The "and / or" operator describes the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: A alone, A and B simultaneously, and B alone. Typically, the character " / " indicates that the objects before and after it are in an "or" relationship. The terms "first," "second," and "third," etc., used in this application are merely for distinguishing similar objects and do not represent a specific ordering of the objects.

[0071] The method embodiments provided in this example can be executed on a terminal, computer, or similar computing device. For example, it can run on a terminal. Figure 1 This is a hardware structure block diagram of the terminal for the training method of the vehicle battery health state estimation model executed in this invention. For example... Figure 1 As shown, a terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 and a memory 104 for storing data are also included. The processor 102 may be, but is not limited to, a microprocessor (MCU) or a programmable logic device (FPGA). The terminal may also include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that… Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the terminal described above. For example, the terminal may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0072] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the training method of the vehicle battery health state estimation model in this invention. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0073] The transmission device 106 is used to receive or send data via a network. This network includes a wireless network provided by the terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 can be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0074] This invention provides a training method for a vehicle battery health state estimation model. Figure 2 This is a flowchart of the training method for the vehicle battery health state estimation model in an embodiment of the present invention, as shown below. Figure 2 As shown, the process includes the following steps:

[0075] Step S210: Obtain the vehicle's battery data; wherein the battery data includes first battery data and second battery data.

[0076] In this step, the model training device first needs to acquire battery data from all vehicles. This battery data can be divided into first battery data and second battery data. The first battery data mainly contains information that reflects the user's driving habits, while the second battery data mainly contains information that reflects the battery's health status.

[0077] Specifically, the first battery data includes at least one of the following: average charging current, average voltage at the end of charging, average state of charge during charging, average charging temperature, charging capacity, average ambient temperature, average speed per 100 kilometers per hour, number of accelerations per 100 kilometers per hour, number of decelerations per 100 kilometers per hour, number of fast charges, number of slow charges, cell type, and total mileage. The second battery data includes at least one of the following: charging current, charging voltage, battery temperature, state of charge, and cumulative charge during slow charging, as well as historical health status values ​​before slow charging.

[0078] The first battery data can include up to 13 different types of attributes, while the second battery data can include up to 6 different types of attributes. The model training device can extract some of the attribute data from the 13 data items as the first battery data, or it can extract some of the attribute data from the 6 data items as the second battery data. In a preferred embodiment, the first battery data includes all 13 different types of attributes, and the second battery data includes all 6 different types of attributes, to facilitate more accurate segmentation of vehicle groups and training of the battery estimation model.

[0079] Step S220: The first battery data is clustered using a preset clustering algorithm to divide the first battery data into multiple categories.

[0080] Specifically, after acquiring the vehicle's battery data, the model training device uses a pre-defined clustering algorithm to cluster the first battery data, thereby classifying it into multiple types. Further, the clustering process classifies multiple first battery data points from each vehicle as a single data sample; that is, each data sample includes first battery data points from multiple different types of attributes for the same vehicle. For example, when the first battery data only includes the average charging current and the average voltage at the end of charging, then a single data sample includes both the average charging current and the average voltage at the end of charging for a particular vehicle.

[0081] For example, the k-means clustering algorithm can be used to cluster the first battery data into k (set value) clusters, or the k-means++ clustering algorithm or the DBSCAN clustering algorithm can be used. Among them, the k-means clustering algorithm and the k-means++ clustering algorithm are distance-based clustering algorithms, while the DBSCAN clustering algorithm is a density-based clustering algorithm.

[0082] Step S230: Divide the vehicles into multiple vehicle groups based on the multi-class first battery data.

[0083] In this step, the model training device divides all vehicles into multiple vehicle groups based on the clustering results of the first battery data. Specifically, data clustering classifies samples containing multiple first battery data points, and each sample corresponds one-to-one with a vehicle. Therefore, the vehicles can be directly divided into multiple vehicle groups (or user groups) based on the clustering results.

[0084] Step S240: For each of the multiple vehicle groups, train the corresponding battery estimation model using the second battery data corresponding to the vehicle group.

[0085] In this step, the model training device trains a battery estimation model for each vehicle group. This means there's a one-to-one correspondence between the battery estimation model and the vehicle group; or rather, each vehicle group has a model for estimating the battery health status of vehicles within its group. Specifically, for each vehicle group, the corresponding battery estimation model is trained using the second battery data for that vehicle group.

[0086] Step S250: Obtain the trained battery estimation model for each vehicle group; wherein, the battery estimation model is used to estimate the battery health status of the vehicles in the corresponding vehicle group.

[0087] In this step, after completing model training, the model training device can obtain the trained battery estimation model for each vehicle group. Therefore, for any vehicle group, all vehicles within that group can input their own second battery data into the battery estimation model corresponding to that vehicle group to estimate their own battery health status.

[0088] The above steps first involve clustering the existing first battery data for all vehicles, and then dividing the vehicles into multiple vehicle groups based on the clustering results. For each vehicle group, a corresponding battery estimation model is trained using the second battery data of that group. Then, all vehicles within a given vehicle group can estimate their own battery health status using the corresponding battery estimation model. Compared to related technologies that configure a separate battery estimation model for each electric vehicle, this significantly reduces the number of battery estimation models, avoids generating a large number of nearly homogeneous models in the battery estimation system, and thus greatly reduces the system's operating load and saves system resources. Therefore, the vehicle battery health status estimation model training method proposed in this invention solves the problem of excessive system operating load and resource waste caused by configuring a separate battery estimation model for each electric vehicle in related technologies.

[0089] Furthermore, the high similarity of user habits within the same vehicle group ensures that the training of the battery estimation model is not affected by data differences. Compared to using a separate battery estimation model for each user's vehicle, the battery estimation model in this invention has more sample data, resulting in a more accurate estimation of battery health after training.

[0090] In some embodiments, step S240, training a battery estimation model using the second battery data corresponding to the vehicle group, specifically includes:

[0091] Step S241: Train multiple autoencoder models based on various second battery data corresponding to the vehicle group.

[0092] Step S242: Obtain the hidden vector corresponding to the second battery data based on the second battery data and its corresponding trained autoencoder model.

[0093] Step S243: Train a self-learning model based on hidden vectors.

[0094] Step S244: Combine the trained autoencoder models and self-learning models into a battery estimation model corresponding to the vehicle group.

[0095] In this embodiment, a relatively specific method for training a battery estimation model is provided. This method can be applied to each vehicle group, meaning that each vehicle group can use the above method steps to train its own battery estimation model.

[0096] Specifically, for any given vehicle group, the first step is to train multiple autoencoder models based on various types of second battery data corresponding to that vehicle group. For example, when the second battery data includes six different types of attributes: charging current, charging voltage, battery temperature, state of charge (SOC), and accumulated charge, as well as historical health status before SOC, then six autoencoder models need to be trained, each corresponding to one of the six types of second battery data. During training, each type of second battery data is input into its corresponding autoencoder model. This results in six trained autoencoder models (each vehicle group has six autoencoder models).

[0097] Furthermore, after the autoencoder model is trained, each second battery data point is input into its respective autoencoder model. The corresponding hidden vector for each second battery data point can then be obtained from the autoencoder model. This means that for each second battery data point, a corresponding hidden vector can be obtained. These hidden vectors can then be used to train a self-learning model (each vehicle group has one self-learning model).

[0098] Finally, the six trained autoencoder models and one self-learning model are combined to form a battery estimation model. Therefore, the estimation process of the battery estimation model is as follows: the second battery data of the vehicle to be estimated is input into the six autoencoder models according to their type attributes, resulting in six corresponding hidden vectors. These six hidden vectors are then input into the self-learning model to finally obtain the vehicle's battery health status estimation result.

[0099] Furthermore, in some embodiments, step S241, training multiple autoencoder models based on various second battery data corresponding to the vehicle group, includes:

[0100] Step S241a: Determine several data samples based on multiple second battery data corresponding to the vehicle group; wherein each data sample includes multiple second battery data generated by the same vehicle in the same charging process.

[0101] Step S241b: Input multiple second battery data from several data samples into multiple autoencoder models to be trained.

[0102] Step S241c: Calculate the first mean squared error loss value between the input and output values ​​of the autoencoder model.

[0103] In step S241d, the model parameters of the corresponding autoencoder model are adjusted according to the first mean squared error loss value until the first mean squared error loss value is less than the first preset threshold, thus completing the training of the autoencoder model.

[0104] In this embodiment, a specific method for training an autoencoder model is further provided.

[0105] Specifically, the second battery data corresponding to the vehicle group is first divided into multiple data samples based on a single charging process of the vehicle. Each data sample includes multiple second battery data generated by the same vehicle in the same charging process. For example, when the second battery data includes six different types of attributes: charging current, charging voltage, battery temperature, state of charge (SCC) value, and cumulative charge amount in slow charging mode, as well as historical health status value before slow charging, then a data sample includes the six data points generated by vehicle A in the nth charging process: charging current, charging voltage, battery temperature, SCC value, and cumulative charge amount in slow charging mode.

[0106] Then, multiple second-cell data points from each data sample are input into their respective autoencoder models to be trained. The first mean squared error loss between the input and output values ​​of the autoencoder model is then calculated. The input values ​​of the autoencoder model are the corresponding second-cell data, and its output value is the data reconstructed from the input second-cell data.

[0107] Finally, with the training principle of minimizing the first mean squared error loss, the parameters of the autoencoder model are gradually adjusted so that the first mean squared error loss of the autoencoder model gradually decreases until the first mean squared error loss is less than the first preset threshold, thus completing the training of the autoencoder model.

[0108] Figure 3 This is a schematic diagram of the structure of the autoencoder model in an embodiment of the present invention. (Refer to...) Figure 3 It should be further explained that the autoencoder model includes an encoding model 310 and a decoding model 320; the hidden vector is the output of the encoding model 310 and the input of the decoding model 320; the input of the encoding model 310 is the input of the autoencoder model, and the output of the decoding model 320 is the output of the autoencoder model.

[0109] Specifically, the autoencoder model mainly consists of two parts: an encoding model 310 and a decoding model 320. When the input vector enters the autoencoder model, it is first processed by the encoding model 310 to obtain an intermediate vector, which is the hidden vector. Then, the intermediate vector is processed by the decoding model 320 to obtain the output vector, thereby realizing vector reconstruction.

[0110] The autoencoder model can be a relatively basic AE model, or a variant of the AE model such as the SAE model, DAE model, or VAE model. Furthermore, the autoencoder model can be constructed using two GRU models (an encoder and a decoder model), or it can be constructed using two other neural network models, such as a CNN model and an LSTM model.

[0111] For example, the training process of the autoencoder model is illustrated below by using charging current data to train an AE model constructed from two GRU models.

[0112] Step a, define the charging current data for each slow charge as a vector cur. i (i represents the i-th slow charge), and each vector contains the current change during this slow charge.

[0113] Step b, change vector cur i As input to the AE model, the AE-based coding model (bidirectional GRU) can convert the input vector cur i This is mapped to a vector z. The specific calculation process is as follows:

[0114]

[0115]

[0116]

[0117] in, and These represent the forward and backward GRU neural network units in a bidirectional GRU, respectively. The difference is that the backward GRU inputs data elements from back to front, and j represents the j-th element of the input vector. Subsequently, vector z maps the decoder part (three-layer GRU) of the input AE into the output vector cur. i，out For the output vector cur i，out The calculation process for the j-th element is as follows:

[0118] l1=g1(z j )

[0119] l2 = g2(l1)

[0120]

[0121] Here, g1, g2, and g3 represent the GRU neural network units in layers 1, 2, and 3, respectively. Therefore, the AE model processes the input vector cur... i The reconstruction was completed, and the step cur was used to obtain an output vector of equal length. i，out .

[0122] Step c: For all input vectors cur, calculate the mean squared error loss between the input and output vectors, and iteratively optimize the AE parameters using the Adam optimizer until the loss is less than the set threshold ε1 = 0.001. The mean squared error loss is calculated using the following formula:

[0123] loss = avg(||cur) i，out -cur i || 2 )

[0124] By following the steps described above, an autoencoder model for the charging current data can be trained. Performing these steps on all second battery data will yield autoencoder models for various second battery data types. Therefore, six autoencoder models with identical structures but different parameters can be trained for six types of second battery data from the same user group.

[0125] Accordingly, in some other embodiments, step S242, obtaining the hidden vector corresponding to the second battery data based on the second battery data and its corresponding trained autoencoder model, includes:

[0126] Step S242a: Input multiple second battery data from several data samples into multiple trained autoencoder models respectively.

[0127] Step S242b: Obtain the hidden vector corresponding to each second battery data from the autoencoder model.

[0128] Specifically, after multiple autoencoder models have been trained, the multiple second battery data points from each data sample are input into the corresponding autoencoder model. The autoencoder model can then obtain the hidden vector corresponding to each second battery data point. For example, when a data sample includes six second battery data points, six hidden vectors can be obtained.

[0129] Step S243, training the self-learning model based on the hidden vectors, includes:

[0130] Step S243a: Determine several feature samples based on the hidden vectors; wherein each feature sample includes multiple hidden vectors corresponding to multiple second battery data in the same data sample.

[0131] Step S243b: Input several feature samples into the self-learning model to be trained.

[0132] Step S243c: Calculate the second mean squared error loss value between the output value of the self-learning model and the corresponding true value of the sample.

[0133] In step S243d, the model parameters of the self-learning model are adjusted according to the second mean squared error loss value until the second mean squared error loss value is less than the second preset threshold, thus completing the training of the self-learning model.

[0134] Specifically, multiple hidden vectors corresponding to each data sample are combined into a single feature sample. For example, when a data sample includes six second battery data points, six hidden vectors can be obtained, which are then combined to form a feature sample. Therefore, each data sample has a corresponding feature sample, and thus, there is a one-to-one correspondence between feature samples and vehicles. The training process of the self-learning model involves inputting the feature sample into the self-learning model and then calculating the mean squared error loss between the model's output value for that feature sample and the corresponding true value. Finally, with the training principle of minimizing the second mean squared error loss, the autoencoder model parameters are gradually adjusted so that the first mean squared error loss of the autoencoder model gradually decreases until the second mean squared error loss is less than a second preset threshold, at which point the training of the self-learning model is complete.

[0135] It should be further noted that the self-learning model can use the TCN model, or other deep learning models or machine learning models, such as the CNN model, LSTM model, XGBoost model and SVM model.

[0136] The following example uses the hidden vectors of six types of second battery data to train a TCN model to illustrate the training process of a self-learning model.

[0137] The six second-cell data points from each data sample are input into the corresponding trained autoencoder model. For any given data sample, the six second-cell data points yield six hidden vectors, which are then combined to form new feature samples. These feature samples and the ground truth values ​​are used to construct a new feature training dataset for subsequent TCN model training. The TCN model training steps include:

[0138] Step a, the feature sample is denoted as f i (i represents the i-th slow charge), and each feature sample corresponds to a specific charging process for a particular vehicle. The feature data samples are input into the TCN model, and the estimated value SOH is output. pre The TCN model structure is shown in the figure below. Feature data is input into two networks consisting of dilated causal convolutional layers, weight normalization layers, ReLU, and Dropout. The output of these networks is summed with the output obtained after the feature data passes through a 1×1 convolutional layer. The sum is used as the final output value SOH. pre

[0139] Step b: For all feature data samples of a certain user group, calculate the corresponding SOH for each sample using TCN. pre Value. Using the estimated SOH pre Value and actual SOH real The MSE of the tags is calculated, and the Adam optimizer is used to adjust the TCN parameters until the MSE of the TCN is less than the set threshold ε2 = 0.01. The MSE calculation process of TCN is as follows:

[0140] loss tcn =avg(||SOH pre -SOH real || 2 )

[0141] This invention also provides a method for estimating the health status of a vehicle battery. Figure 4 This is a flowchart of a method for estimating the health status of a vehicle battery according to an embodiment of the present invention, such as... Figure 4 As shown, the process includes the following steps:

[0142] Step 410: Obtain the battery data of the vehicle to be estimated; wherein the battery data includes first battery data and second battery data.

[0143] Step 420: Determine the vehicle group to which the vehicle to be estimated belongs based on the first battery data, and determine the battery estimation model corresponding to the vehicle to be estimated based on the vehicle group.

[0144] Step 430: Determine the battery health status of the vehicle to be estimated using the battery estimation model and the second battery data.

[0145] The battery estimation model is trained using the training method for the vehicle battery health state estimation model provided in this invention.

[0146] Specifically, the process begins by acquiring battery data for the vehicle to be estimated, including first and second battery data. Then, based on the first battery data, the vehicle group to which the vehicle belongs is determined, thus identifying the appropriate battery estimation model. Finally, the second battery data is input into the corresponding battery estimation model to obtain an estimate of the vehicle's battery health status. The battery estimation model is trained using the training method described in this application for the vehicle battery health status estimation model.

[0147] The technical solutions of the present invention will now be described and illustrated through specific preferred embodiments.

[0148] The vehicle battery health state estimation method in this preferred embodiment divides users with similar driving habits into a group based on user profile data, and learns a low-dimensional representation of vehicle data characteristics in the same user group by leveraging the powerful nonlinear expression capability of AE. Finally, it uses an efficient TCN model to complete multi-feature fusion to accurately estimate SOH (battery health state).

[0149] Figure 5 This is a flowchart of a method for estimating the health status of a vehicle battery according to a preferred embodiment of the present invention.

[0150] Reference Figure 5 The process includes the following steps:

[0151] Step S510: Obtain the vehicle's first battery data.

[0152] Specifically, firstly, relevant signal data such as the user's vehicle power battery are uploaded to the power battery data cloud platform via communication equipment for unified storage and management. Then, the battery estimation device extracts 13 attribute data points for each user from the battery data cloud platform: average charging current, average voltage at the end of charging, average SOC (state of charge), average charging temperature, charging capacity, average ambient temperature, average speed per 100 kilometers, number of acceleration / deceleration cycles per 100 kilometers, number of fast / slow charging cycles, cell type, and total mileage, thereby constructing a user profile for each user.

[0153] Step S520: The first battery data is clustered using a preset clustering algorithm, and the first battery data is divided into multiple categories. Based on the multiple categories of first battery data, the vehicles are divided into multiple vehicle groups.

[0154] Specifically, the k-means method is used to cluster the aforementioned user profile data (first battery data) into k (set value) clusters, and the users in each cluster are defined as a group. Thus, all users can be divided into k different user groups (corresponding to vehicle groups). The detailed clustering process includes the following steps:

[0155] Step a1: Each user's profile data can be represented as a sample point u. vin (A vector composed of 13 profile data items, where vin is the user's vehicle identification number), from all u vin Randomly select k sample points as the initial cluster centers.

[0156] Step a2: Calculate the Euclidean distance between each remaining sample and the k cluster centers, and assign each sample to the cluster containing the nearest cluster center. The Euclidean distance formula is as follows:

[0157]

[0158] Step a3: Calculate the mean of all sample point vectors in each of the k clusters, and use the k mean vectors as the new cluster centers.

[0159] Step a4: Repeat steps a2 and a3 until the k mean vectors are no longer updated. At this point, the users in the k clusters form k user groups.

[0160] Step S530: For each vehicle group, obtain its corresponding second battery data.

[0161] Specifically, based on the identified k user groups, battery-related data (second battery data) of user vehicles in each group is retrieved from the power battery data cloud platform, and k training datasets are constructed by combining them with real SOH value labels. The battery-related data mainly includes the following six items: charging current, charging voltage, battery temperature, SOC value, and cumulative charging amount under slow charging conditions with a sampling interval of 3 minutes, and historical SOH value before slow charging.

[0162] Step S540: For each vehicle group, train multiple autoencoder models based on the corresponding second battery data.

[0163] Specifically, for any user group, the AE (mean squared error loss) model is trained using data from the six battery types in their training dataset until the MSE of all six AEs is less than the threshold ε1. The specific training process (the training process is consistent for all six battery types; here, we only use charging current data as an example) includes the following steps:

[0164] Step b1: Define the charging current data for each slow charge as a vector cur.i (i represents the i-th slow charge), and each vector contains the current change during this slow charge.

[0165] Step b2, change the vector cur i As input to the AE model, the AE-based coding model (bidirectional GRU) can convert the input vector cur i This is mapped to a vector z. The specific calculation process is as follows:

[0166]

[0167]

[0168]

[0169] in, and These represent the forward and backward GRU neural network units in a bidirectional GRU, respectively. The difference is that the backward GRU inputs data elements from back to front, and j represents the j-th element of the input vector. Subsequently, vector z maps the decoder part (three-layer GRU) of the input AE into the output vector cur. i，out For the output vector cur i，out The calculation process for the j-th element is as follows:

[0170] l1=g1(z j )

[0171] l2 = g2(l1)

[0172]

[0173] Here, g1, g2, and g3 represent the GRU neural network units in layers 1, 2, and 3, respectively. Therefore, the AE model processes the input vector cur... i The reconstruction was completed, and the step cur was used to obtain an output vector of equal length. i，out .

[0174] Step b3: For all input vectors cur, calculate the mean squared error loss between the input and output vectors, and iteratively optimize the AE parameters using the Adam optimizer until the loss is less than the set threshold ε1 = 0.001. The mean squared error loss is calculated using the following formula:

[0175] loss = avg(||cur) i，out -cur i || 2 )

[0176] Step S550: Obtain the hidden vector corresponding to the second battery data based on the second battery data and its corresponding trained autoencoder model.

[0177] Specifically, after each AE (Advanced Elementary) training is completed, the training dataset for that user group is re-inputted into the AE, and the six hidden variables corresponding to each training sample are merged to form a new feature training sample. Combined with the corresponding true SOH (Solar Oxygen Response) value labels for each sample, a feature training dataset is constructed. Further, the six types of battery data from the original data are input into their respective AE models. For any training data sample, the six original battery data it contains can yield six vectors z, which are merged to form a new feature data sample. Using these feature data samples and the true SOH value labels, a new feature training dataset can be constructed to complete the subsequent training of the TCN (Training Channel Network) model.

[0178] Step S560: Train a self-learning model based on hidden vectors.

[0179] Specifically, through the above steps, a feature training dataset for this user group can be established and used to train the TCN model until the MSE of the TCN model is less than the threshold ε2. The specific training process includes the following steps:

[0180] Step c1, the feature sample is denoted as f i (i represents the i-th slow charge), and each feature sample corresponds to a specific charging process for a particular vehicle. The feature data samples are input into the TCN model, and the estimated value SOH is output. pre The TCN model structure is shown in the figure below. Feature data is input into two networks consisting of dilated causal convolutional layers, weight normalization layers, ReLU, and Dropout. The output of these networks is summed with the output obtained after the feature data passes through a 1×1 convolutional layer. The sum is used as the final output value SOH. pre

[0181] Step c2: For all feature data samples of a certain user group, calculate the corresponding SOH for each sample using TCN. pre Value. Using the estimated SOH pre Value and actual SOH real The MSE of the tags is calculated, and the Adam optimizer is used to adjust the TCN parameters until the MSE of the TCN is less than the set threshold ε2 = 0.01. The MSE calculation process of TCN is as follows:

[0182] loss tcn =avg(||SOH pre -SOH real || 2 )

[0183] Step S570: Combine the trained autoencoder models and self-learning models into a battery estimation model corresponding to the vehicle group.

[0184] Specifically, for k user groups, the above steps are performed using their respective training datasets, thereby establishing k SOH estimation models composed of AE and TCN.

[0185] Step S580: Estimate the battery health status of vehicles within their respective user groups using a battery estimation model.

[0186] Specifically, once all SOH estimation models are established, they can be used for testing. First, user profile data is extracted, and Euclidean distance is used to determine the user's group. Finally, the user's vehicle's slow-charging battery data is input into the SOH estimation model corresponding to the user group to obtain the SOH value. The specific estimation process includes the following two steps:

[0187] Step d1: For users not included in the training phase, extract the vector containing 13 profile data points for that user from the cloud platform, and calculate its Euclidean distance to the cluster centers of each user group. Assign the user to the nearest user group.

[0188] Step d2: Call the SOH estimation model of the user's group and extract 6 battery data points from the user's vehicle during the most recent slow charging through the cloud platform. Input the data into the SOH estimation model to obtain the SOH value of the user's vehicle.

[0189] The vehicle battery health status estimation method in the above preferred embodiments has the following advantages:

[0190] 1. Currently, many SOH estimation methods use the same estimation model for all EVs, or design a separate estimation model for each EV. However, this invention designs an SOH estimation model based on user groups. This not only avoids the model performance degradation problem caused by EV data differences, but also allows user vehicles with similar characteristics to share a single SOH estimation model, significantly reducing the number of models and lowering the burden on the cloud platform.

[0191] 2. The input data is reconstructed using a neural network (AE), enabling the AE to obtain low-dimensional representations of various battery data features. Inputting these features into the TCN model not only helps the TCN model to more comprehensively understand the current state of the user's vehicle and accurately estimate SOH, but also avoids repeated debugging, comparison, and modification of the feature extraction method in the early stages of model construction.

[0192] This embodiment also provides a training device for a vehicle battery health state estimation model and a vehicle battery health state estimation device. This device is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. The terms "module," "unit," "subunit," etc., used below refer to combinations of software and / or hardware that implement a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementations, or combinations of software and hardware, are also possible and contemplated.

[0193] Figure 6 This is a structural block diagram of the training device for the vehicle battery health state estimation model in an embodiment of the present invention, as shown below. Figure 6 As shown, the device includes:

[0194] The data acquisition module 610 is used to acquire battery data of the vehicle; wherein the battery data includes first battery data and second battery data.

[0195] The data clustering module 620 is used to cluster the first battery data using a preset clustering algorithm, thereby dividing the first battery data into multiple categories;

[0196] The group segmentation module 630 is used to divide the vehicles into multiple vehicle groups based on multiple types of first battery data;

[0197] The model training module 640 is used to train a battery estimation model for each of the multiple vehicle groups using the second battery data corresponding to the vehicle group.

[0198] The model determination module 650 is used to obtain the trained battery estimation model for each vehicle group; wherein the battery estimation model is used to estimate the battery health status of the vehicles in the corresponding vehicle group.

[0199] Figure 7 This is a structural block diagram of a vehicle battery health status estimation device according to an embodiment of the present invention, as shown below. Figure 7 As shown, the device includes:

[0200] The data acquisition module 710 is used to acquire battery data of the vehicle to be estimated; wherein, the battery data includes first battery data and second battery data;

[0201] The model determination module 720 is used to determine the vehicle group to which the vehicle to be estimated belongs based on the first battery data, and to determine the battery estimation model corresponding to the vehicle to be estimated based on the vehicle group.

[0202] The state assessment module 730 is used to determine the battery health status of the vehicle to be estimated through the battery estimation model and the second battery data.

[0203] The battery estimation model is trained using the training device for the vehicle battery health state estimation model provided in this invention.

[0204] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0205] This embodiment also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0206] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated in this embodiment.

[0207] Furthermore, in conjunction with the training method for the vehicle battery health state estimation model provided in the above embodiments, this embodiment can also provide a storage medium for implementation. This storage medium stores a computer program; when executed by a processor, the computer program implements any of the training methods for the vehicle battery health state estimation model in the above embodiments.

[0208] Furthermore, in conjunction with the vehicle battery health state estimation methods provided in the above embodiments, this embodiment can also provide a storage medium for implementation. This storage medium stores a computer program; when executed by a processor, the computer program implements any of the vehicle battery health state estimation methods described in the above embodiments.

[0209] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0210] It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. All other embodiments derived by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0211] Obviously, the accompanying drawings are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar situations based on these drawings without any creative effort. Furthermore, it is understood that although the work done in this development process may be complex and lengthy, for those skilled in the art, certain design, manufacturing, or production modifications made based on the technical content disclosed in this application are merely conventional technical means and should not be considered as insufficient disclosure of this application.

[0212] The term "embodiment" in this application refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily imply the same embodiment, nor does it imply that it is mutually exclusive with or independent of other embodiments. It will be clearly or implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0213] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A training method for a vehicle battery health state estimation model, characterized in that, The method includes: Obtain vehicle battery data; wherein the battery data includes first battery data and second battery data; A preset clustering algorithm is used to cluster the first battery data, dividing the first battery data into multiple categories; The vehicles are divided into multiple vehicle groups based on the first battery data of multiple categories; For each of the multiple vehicle groups, a battery estimation model corresponding to the vehicle group is trained using the second battery data corresponding to the vehicle group. The step of training a battery estimation model using the second battery data corresponding to the vehicle group includes: training multiple autoencoder models based on various types of second battery data corresponding to the vehicle group; obtaining a hidden vector corresponding to the second battery data based on the second battery data and the corresponding trained autoencoder models; training a self-learning model based on the hidden vector; and combining the trained multiple autoencoder models and the self-learning model into a battery estimation model corresponding to the vehicle group. A trained battery estimation model is obtained for each vehicle group; wherein the battery estimation model is used to estimate the battery health status of the vehicles in the corresponding vehicle group.

2. The training method for the vehicle battery health state estimation model according to claim 1, characterized in that, The process of training multiple autoencoder models based on various second battery data corresponding to the vehicle group includes: Several data samples are determined based on various second battery data corresponding to the vehicle group; wherein each data sample includes multiple second battery data generated by the same vehicle during the same charging process; Multiple second battery data from several data samples are respectively input into multiple autoencoder models to be trained; Calculate the first mean squared error loss value between the input and output values ​​of the autoencoder model; The model parameters of the autoencoder model are adjusted according to the first mean squared error loss value until the first mean squared error loss value is less than a first preset threshold, thus completing the training of the autoencoder model.

3. The training method for the vehicle battery health state estimation model according to claim 2, characterized in that, The step of obtaining the hidden vector corresponding to the second battery data based on the second battery data and its corresponding trained autoencoder model includes: Multiple second battery data from several of the data samples are respectively input into multiple trained autoencoder models; Obtain the hidden vector corresponding to each of the second battery data from the autoencoder model; The self-learning model trained based on the hidden vectors includes: Several feature samples are determined based on the hidden vectors; wherein each feature sample includes multiple hidden vectors corresponding to multiple second battery data in the same data sample; Several of the aforementioned feature samples are input into the self-learning model to be trained; Calculate the second mean squared error loss value between the output value of the self-learning model and the corresponding true value of the sample; The model parameters of the self-learning model are adjusted according to the second mean squared error loss value until the second mean squared error loss value is less than the second preset threshold, thus completing the training of the self-learning model.

4. The training method for the vehicle battery health state estimation model according to claim 1, characterized in that, The autoencoder model includes an encoding model and a decoding model; The hidden vector is the output of the encoding model and the input of the decoding model; The input to the encoding model is the input to the autoencoder model, and the output of the decoding model is the output of the autoencoder model.

5. The training method for the vehicle battery health state estimation model according to any one of claims 1 to 4, characterized in that, The first battery data includes at least one of the following: Average charging current, average voltage at the end of charging, average state of charge, average charging temperature, charging capacity, average ambient temperature, average speed per 100 kilometers, number of accelerations per 100 kilometers, number of decelerations per 100 kilometers, number of fast charging cycles, number of slow charging cycles, cell type, and total mileage. The second battery data includes at least one of the following: The charging current, charging voltage, battery temperature, state of charge (SOC) value, and cumulative charge amount during slow charging, as well as the historical health status values ​​before slow charging.

6. A method for estimating the state of health of a vehicle battery, characterized in that, The method includes: Obtain battery data of the vehicle to be estimated; wherein the battery data includes first battery data and second battery data; The vehicle group to which the vehicle to be estimated belongs is determined based on the first battery data, and the battery estimation model corresponding to the vehicle to be estimated is determined based on the vehicle group. The battery health status of the vehicle to be estimated is determined using the battery estimation model and the second battery data. The battery estimation model is trained using the method described in any one of claims 1 to 5.

7. A training device for a vehicle battery health state estimation model, characterized in that, The device includes: A data acquisition module is used to acquire battery data of the vehicle; wherein the battery data includes first battery data and second battery data; The data clustering module is used to cluster the first battery data using a preset clustering algorithm, thereby dividing the first battery data into multiple categories; A group segmentation module is used to divide the vehicle into multiple vehicle groups based on multiple types of the first battery data; The model training module is used to train a battery estimation model corresponding to each of the multiple vehicle groups using the second battery data corresponding to the vehicle group. The model training module is also used to train multiple autoencoder models based on various second battery data corresponding to the vehicle group; obtain hidden vectors corresponding to the second battery data according to the second battery data and the corresponding trained autoencoder models; train a self-learning model based on the hidden vectors; and combine the trained multiple autoencoder models and the self-learning model into a battery estimation model corresponding to the vehicle group. The model determination module is used to obtain a trained battery estimation model for each vehicle group; wherein the battery estimation model is used to estimate the battery health status of the vehicles in the corresponding vehicle group.

8. A device for estimating the health status of a vehicle battery, characterized in that, The device includes: A data acquisition module is used to acquire battery data of the vehicle to be estimated; wherein the battery data includes first battery data and second battery data; The model determination module is used to determine the vehicle group to which the vehicle to be estimated belongs based on the first battery data, and to determine the battery estimation model corresponding to the vehicle to be estimated based on the vehicle group. A state assessment module is used to determine the battery health status of the vehicle to be estimated using the battery estimation model and the second battery data. The battery estimation model is trained using the apparatus described in claim 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the training method for the vehicle battery health state estimation model according to any one of claims 1 to 5, or the steps of the vehicle battery health state estimation method according to claim 6.

Images