A method, device and related products for establishing a battery pack life attenuation model

Abstract

The present application discloses a method, device and related products for establishing a battery pack life decay model. A life decay basic model of a target battery pack is established based on usage data and life decay data of a reference battery pack; target battery pack usage data is obtained; transfer learning is performed based on the target battery pack usage data and the reference battery pack usage data to obtain a mapping relationship between the target battery pack usage data and the reference battery pack usage data; the life decay model of the target battery pack is obtained based on the mapping relationship and the life decay basic model. The target battery pack life decay basic model is established based on the reference battery pack usage data and life decay data. By establishing a mapping relationship between the target battery pack usage data and the reference battery pack usage data, the matching degree of the model can be improved when using the life decay basic model, so that the obtained life decay model can better serve the life decay prediction for the target battery pack.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a method and device for establishing a battery pack life attenuation model and related products. Background Art

[0002] Before a new electric vehicle is launched on the market, there is usually a lack of life decay data for the battery pack of the electric vehicle. Even if the life decay data is obtained experimentally, on the one hand, it consumes experimental costs, and on the other hand, the battery pack used for the experiment cannot be used in the newly launched electric vehicles. Due to the lack of battery pack life decay data and cost reasons, it is currently difficult to establish a battery pack life decay model, so it is impossible to accurately predict the life decay of the battery pack of the newly launched electric vehicle. For the battery packs of such newly launched electric vehicles, how to establish an accurate battery pack life decay model has become a technical problem that needs to be solved urgently in this field. Summary of the invention

[0003] Based on the above problems, the present application provides a method, device and related products for establishing a battery pack life decay model to improve the accuracy of the established life decay model.

[0004] The embodiments of the present application disclose the following technical solutions:

[0005] In a first aspect, the present application provides a method for establishing a battery pack life attenuation model, comprising:

[0006] Establish a life decay basic model for the target battery pack based on the usage data and life decay data of the reference battery pack;

[0007] Obtaining usage data of the target battery pack;

[0008] Performing transfer learning according to the usage data of the target battery pack and the usage data of the reference battery pack to obtain a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack;

[0009] The life decay model of the target battery pack is obtained according to the mapping relationship and the life decay basic model.

[0010] Optionally, the above method further includes:

[0011] When the training samples of the target battery pack increase, the life decay model of the target battery pack is updated through an incremental learning mode; the training samples include: usage data of the target battery pack and life decay data corresponding to the usage data.

[0012] Optionally, the life decay model of the target battery pack is established by using a support vector machine; and updating the life decay model of the target battery pack by using an incremental learning mode includes:

[0013] Using the support vector machine to learn the first batch of training samples to obtain a first batch of support vectors; the first batch of support vectors are training samples selected by the support vector machine from the first batch of training samples;

[0014] The support vector machine is trained again using a second batch of training samples and the first batch of support vectors to update the life decay model of the target battery pack; the second batch of training samples are training samples of the target battery pack that are newly added relative to the first batch of training samples.

[0015] Optionally, the life decay model includes a fixed coefficient layer and a variable coefficient layer;

[0016] The updating of the life attenuation model of the target battery pack by the incremental learning mode includes:

[0017] The life decay model is updated according to the newly added training samples of the target battery pack, and the update only changes the coefficients of the coefficient variable layer.

[0018] Optionally, the reference battery pack is a battery pack of the same type as the target battery pack, or the reference battery pack is a battery pack of a similar type as the target battery pack.

[0019] In a second aspect, the present application provides a device for establishing a battery pack life attenuation model, comprising:

[0020] A basic model building module, used to build a life decay basic model of a target battery pack based on usage data and life decay data of a reference battery pack;

[0021] A target battery pack data acquisition module, used to obtain usage data of the target battery pack;

[0022] a transfer learning module, configured to perform transfer learning according to the usage data of the target battery pack and the usage data of the reference battery pack, and obtain a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack;

[0023] A life decay model establishment module is used to obtain the life decay model of the target battery pack according to the mapping relationship and the life decay basic model.

[0024] Optionally, the above device also includes:

[0025] An updating module is used to update the life decay model of the target battery pack through an incremental learning mode when the training samples of the target battery pack increase; the training samples include: usage data of the target battery pack and life decay data corresponding to the usage data.

[0026] Optionally, the life attenuation model of the target battery pack is established using a support vector machine; the update module includes:

[0027] A first acquisition unit is used to learn the first batch of training samples using the support vector machine to obtain a first batch of support vectors; the first batch of support vectors are training samples selected by the support vector machine from the first batch of training samples;

[0028] The first repeated training unit is used to re-train the support vector machine using a second batch of training samples and the first batch of support vectors to update the life decay model of the target battery pack; the second batch of training samples are training samples of the target battery pack that are newly added relative to the first batch of training samples.

[0029] In a third aspect, the present application provides a device for establishing a battery pack life decay model, comprising a processor and a memory; the memory is used to store a computer program; the processor is used to execute a method for establishing a battery pack life decay model as provided in the first aspect according to the computer program.

[0030] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, executes the method for establishing a battery pack life attenuation model as provided in the first aspect.

[0031] Compared with the prior art, this application has the following beneficial effects:

[0032] In the method for establishing a battery pack life decay model provided in the present application, first, a life decay basic model of a target battery pack is established based on the usage data and life decay data of a reference battery pack; the usage data of the target battery pack is obtained; thereafter, transfer learning is performed based on the usage data of the target battery pack and the usage data of the reference battery pack to obtain a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack; and then the life decay model of the target battery pack is obtained based on the mapping relationship and the life decay basic model. The life decay basic model of the target battery pack is established based on the usage data and life decay data of the reference battery pack, and thus the matching degree with the target battery pack needs to be improved. By establishing a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack, the matching degree of the model can be improved when using the life decay basic model, thereby enhancing the usability, so that the obtained life decay model can better serve the life decay prediction for the target battery pack. Therefore, the application of the technical solution of the present application effectively improves the accuracy of the established life decay model. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0034] Figure 1 A flowchart of a method for establishing a battery pack life attenuation model provided in an embodiment of the present application;

[0035] Figure 2 A flowchart of another method for establishing a battery pack life attenuation model provided in an embodiment of the present application;

[0036] Figure 3 A schematic diagram of the structure of a device for establishing a battery pack life attenuation model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0037] As described above, it is difficult to establish an accurate battery pack life decay model for new electric vehicles due to the lack of battery pack life decay data. If the life decay data is obtained through experiments, there will be cost issues due to experimental losses. Under the conditions of lack of data and cost constraints, how to establish an accurate battery pack life decay model has become a major problem.

[0038] Based on the above problems, the inventors have provided a method, device and related products for establishing a battery pack life decay model after research. In the technical solution of the present application, a basic model is first established as the life decay basic model of the target battery pack using readily available and sufficient usage data and life decay data of the reference battery pack. And the mapping relationship between the target battery pack and the reference battery pack in the usage data is obtained through transfer learning. Finally, a relatively more accurate life decay model for the target battery pack is obtained by using the above mapping relationship obtained through transfer learning and the life decay basic model established previously. This life decay model can achieve a more accurate life decay prediction for the target battery pack. Thereby solving the above problems.

[0039] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0040] Method Embodiment

[0041] See also Figure 1 , which is a flow chart of a method for establishing a battery pack life attenuation model provided in an embodiment of the present application. Figure 1 As shown, the method for establishing the battery pack life attenuation model includes:

[0042] S101: Establishing a life decay basic model of a target battery pack based on usage data and life decay data of a reference battery pack.

[0043] In the embodiment of the present application, the reference battery pack is a battery pack of other electric vehicles (from the same or different car manufacturers as the electric vehicle to which the target battery pack belongs) that were launched before the electric vehicle to which the target battery pack belongs. For example, the battery pack of the first model of electric vehicles launched by a certain car company in May 2020 is used as the reference battery pack of the second model of electric vehicles launched in August 2020 (the target battery pack).

[0044] In one possible implementation, the reference battery pack is a battery pack of the same type as the target battery pack, or the reference battery pack is a battery pack of a similar type to the target battery pack. The same or similar types here can be defined in many ways. For example, battery packs produced by the same battery cell factory are regarded as battery packs of the same type; battery packs of the same material system (such as lithium iron phosphate) are regarded as battery packs of the same type; battery packs with the same initial capacity or initial resistance are regarded as battery packs of the same type; battery packs with similar structures and material parameters are regarded as similar battery packs, and so on. There is no limitation here.

[0045] Since the electric vehicles to which the reference battery pack belongs were launched earlier, the big data platform can collect or obtain relatively rich and sufficient usage data and life attenuation data of the reference battery pack. There is no limit on the number of reference battery packs.

[0046] The usage data of the reference battery pack may be directly obtained through the battery management system or sensor (such as temperature, current, voltage, etc.), or may be obtained by calculation or model prediction or analysis based on the data obtained by the battery management system or sensor (such as internal resistance). There is no restriction on the content of the usage data.

[0047] The life decay data of the reference battery pack may be obtained based on the usage data of the reference battery pack and the usage data of the vehicle to which the reference battery pack belongs (such as mileage, speed, driving time, etc.), or may be predicted by the life decay model of the reference battery pack, or may be collected through a limited number of experiments. There is no restriction on the way to obtain the reference battery pack.

[0048] The life decay data may include: the change in the battery's state of health (SOH). SOH can be defined in different ways according to specific needs. For example, if SOH is defined by the total capacity of the battery, then SOH is defined as the percentage of the total capacity of the battery to the total capacity of the new battery; if SOH is defined by the internal resistance of the battery, then the relative increase in the internal resistance of the battery compared to the internal resistance of a new battery can be taken as the SOH of the battery. In this step, a training sample can be constructed based on the usage data and life decay data of the reference battery pack, and the training sample can be used to train the basic life decay model of the target battery pack. For example, the usage data of the reference battery pack within 5km mileage is used as the input data of the model training process, and the change in SOH within the corresponding mileage is used as the label of the input data. The basic life decay model obtained after training can realize the function of predicting and evaluating the life decay of the battery pack. Since the life decay basic model is trained based on the usage data and life decay data of the reference battery pack as samples, the trained life decay basic model can make a relatively accurate life decay prediction and evaluation of the reference battery pack. However, since the target battery pack and the reference battery pack may differ due to factors such as the vehicle model, driving environment, and battery cells used, the accuracy of the life decay basic model for the life decay prediction and evaluation of the target battery pack is poor and needs to be improved. To this end, the following S102-S103 needs to be executed.

[0049] S102: Obtain usage data of a target battery pack.

[0050] After the electric vehicle containing the target battery pack is put into use after it is put into use, the usage data of the target battery pack can be obtained. The method of obtaining the usage data of the target battery pack is similar to the method of obtaining the usage data of the reference battery pack, and will not be repeated here.

[0051] S103: Perform transfer learning according to the usage data of the target battery pack and the usage data of the reference battery pack to obtain a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack.

[0052] To perform the operation described in this step, as a possible implementation method, corresponding features can be extracted from the usage data of the target battery pack and the usage data of the reference battery pack, and transfer learning can be performed based on the extracted corresponding features to obtain a mapping relationship between the corresponding features.

[0053] The mapping relationship between the basic battery pack usage data and the target battery pack usage data here is a category of transfer learning: feature-based transfer. Essentially, a regression model is a mapping relationship between input features (X) and labels (y), or it can be understood as a conditional probability distribution p(y|X). In transfer learning, the data of the basic model is called the source domain (marked as s), and the data of the target model to be established is called the target domain (marked as t). For some problems, the label distribution of the source domain and the target domain is similar, and p(y|X) is also similar, but the feature distribution P(Xs) is different from P(Xt). However, some methods can be used to find a mapping f so that the distribution of the source domain and the target domain P(f(X)) is similar, then P(ys|f(Xs)) and P(yt|f(Xt)) are also similar. In this case, if the mapping f can be found, then the model trained based on the source domain can be directly used on the target domain. However, if the label distribution or p(y|X) between the source domain and the target domain is somewhat different (but has the same parts), then when migrating the model of the source domain to the target domain, it needs to be trained, but the migration model can reduce the demand for training samples in disguise.

[0054] Extracting features based on two parts of data and establishing a mapping relationship between the features is a relatively mature technology in this field, so the implementation process of this step is not described in detail here.

[0055] S104: Obtaining a life decay model of the target battery pack according to the mapping relationship and the life decay basic model.

[0056] Compared with the usage data of the target battery pack, the life decay basic model established in S101 is more capable of realizing accurate life decay prediction and evaluation functions based on the usage data of the reference battery pack. In order to accurately predict and evaluate the life decay of the target battery pack based on the input data after the usage data of the target battery pack is input, the mapping relationship obtained in S103 is combined with the life decay basic model obtained in S101 in this step. The life decay basic model essentially extracts the features of the input data, and then predicts and evaluates the changes in the life decay of the battery pack based on the extracted features. By combining the life decay basic model with the above-mentioned mapping relationship, the adjusted model can be adapted to the input usage data of the target battery pack, so that accurate life decay prediction and evaluation can be realized based on the input usage data of the target battery pack.

[0057] The above is the life decay model provided in the embodiment of the present application. The life decay basic model of the target battery pack is established based on the usage data and life decay data of the reference battery pack, and therefore the matching degree with the target battery pack needs to be improved. By establishing a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack, the matching degree of the model can be improved when using the life decay basic model, thereby enhancing the usability, so that the obtained life decay model can better serve the life decay prediction for the target battery pack. Therefore, the application of the technical solution of the present application effectively improves the accuracy of the established life decay model.

[0058] In practical applications, as newly launched electric vehicles gradually become more and more popular, more and more usage data and life decay data of target battery packs can be collected. These usage data and life decay data can be used to establish new training samples to improve the life decay model obtained in the above S104. When new training samples are added, a common operation is to mix the new training samples with the old training samples, and comprehensively use the original old training samples and the new training samples to train and update the model. However, this operation method is easy to waste processor time and has low training efficiency.

[0059] In order to improve the training and updating efficiency of the model and save processor processing time, the present application also provides the following embodiments. Figure 2 A flowchart of another method for establishing a battery pack life attenuation model provided in an embodiment of the present application. Figure 2 As shown, the method includes:

[0060] S201 to S204. The implementation of S201 to S204 is substantially the same as the implementation of S101 to S104 in the aforementioned embodiment, so the relevant description can refer to the aforementioned embodiment and will not be repeated here.

[0061] S205: Construct a new training sample of the target battery pack according to the usage data of the target battery pack and the life attenuation data corresponding to the usage data.

[0062] As an example, a vehicle was launched on August 1, 2020. The usage data of the target battery pack was collected in the first mileage on August 31, 2020, and the SOH change of the target battery pack after the end of the mileage compared with that before the mileage was used as the life decay data corresponding to the usage data of the target battery pack. The above usage data and life decay data can be used as a set of training samples.

[0063] S206: Based on the newly added training samples of the target battery pack, the life attenuation model of the target battery pack is updated through an incremental learning mode.

[0064] Two example implementations of this step are provided below.

[0065] The first implementation method:

[0066] In this implementation, the life attenuation model of the target battery pack is specifically established using a support vector machine. Assuming that the second batch of training samples are newly added training samples compared to the first batch of training samples, this step may specifically include:

[0067] The first batch of training samples are studied using a support vector machine to obtain the first batch of support vectors. When the first batch of support vectors are obtained, the training of the life decay model with the first batch of training samples is completed. The mechanism of the support vector machine is to output support vectors. The first batch of support vectors are the training samples selected by the support vector machine from the first batch of training samples. After obtaining a new batch of training samples (i.e., the second batch of training samples), unlike the method of combining the first batch of training samples with the second batch of training samples, only the second batch of training samples and the first batch of support vectors obtained previously can be used to train the support vector machine again to update the life decay model of the target battery pack. In this way, the number of training samples input is reduced, the processing time of the processor is saved, and the efficiency of model training is improved.

[0068] As an example, the first batch of training samples includes 100 training samples A1-A100. The second batch of training samples includes 50 training samples B1-B50. The first support vector includes training samples A21-A30. When the model is trained again, only training samples B1-B50 and training samples A21-A30 are used.

[0069] The second implementation method:

[0070] Generally speaking, the coefficients of the layers of the model are all variable. The life decay model usually includes multiple layers. In the embodiment of the present application, the coefficients of one or more layers are set to be immutable, and such layers are called fixed coefficient layers. The coefficients of other layers are kept variable, and such layers are called variable coefficient layers.

[0071] In this step, when using the newly added training samples to train and update the life decay model, only the coefficients of the variable coefficient layer are allowed to be changed. In this way, the training time of the model is saved, the training efficiency is improved, and the accuracy of the battery pack life decay model is improved with the newly added training samples, thus improving the model function.

[0072] In combination with the embodiments described above, in the technical solution of the present application, when new training samples are added, the battery pack life decay model is updated by incremental learning, which can significantly improve the training efficiency of the model, save processor processing time, improve the model function, and improve the accuracy of the model in predicting and evaluating life decay changes.

[0073] Based on the method for establishing a battery pack life attenuation model provided in the above-mentioned embodiment, the present application also provides a device for establishing a battery pack life attenuation model. The specific implementation of the device is described below in conjunction with the embodiments and drawings.

[0074] Device Embodiment

[0075] Figure 3 This is a schematic diagram of a device for establishing a battery pack life attenuation model provided in an embodiment of the present application. Figure 3 As shown, the device 300 for establishing the battery pack life attenuation model includes:

[0076] A basic model building module 301 is used to build a life decay basic model of a target battery pack according to usage data and life decay data of a reference battery pack;

[0077] A target battery pack data acquisition module 302 is used to obtain usage data of the target battery pack;

[0078] A transfer learning module 303, configured to perform transfer learning according to the usage data of the target battery pack and the usage data of the reference battery pack, and obtain a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack;

[0079] The life decay model establishing module 304 is used to obtain the life decay model of the target battery pack according to the mapping relationship and the life decay basic model.

[0080] The life decay basic model of the target battery pack is established based on the usage data and life decay data of the reference battery pack, so the matching degree with the target battery pack needs to be improved. By establishing a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack, the matching degree of the model can be improved when using the life decay basic model, thereby enhancing the usability, so that the obtained life decay model can better serve the life decay prediction for the target battery pack. Therefore, the device 300 for establishing the battery pack life decay model provided by the technical solution of the present application effectively improves the accuracy of the established life decay model.

[0081] In order to improve the accuracy of the battery pack life attenuation model, after the model is established, it can be retrained using newly obtained training samples to update the coefficients. However, conventional training methods may have problems such as low training efficiency and long processor processing time.

[0082] Based on the above update problem, optionally, Figure 3 As shown, the device 300 for establishing the battery pack life attenuation model may also include:

[0083] The updating module 305 is used to update the life decay model of the target battery pack through an incremental learning mode when the training samples of the target battery pack increase; the training samples include: usage data of the target battery pack and life decay data corresponding to the usage data.

[0084] In a possible implementation, the life attenuation model of the target battery pack is established using a support vector machine; the update module 305 includes:

[0085] A first acquisition unit is used to learn the first batch of training samples using the support vector machine to obtain a first batch of support vectors; the first batch of support vectors are training samples selected by the support vector machine from the first batch of training samples;

[0086] The first repeated training unit is used to re-train the support vector machine using a second batch of training samples and the first batch of support vectors to update the life decay model of the target battery pack; the second batch of training samples are training samples of the target battery pack that are newly added relative to the first batch of training samples.

[0087] In another possible implementation, the life decay model includes a fixed coefficient layer and a variable coefficient layer;

[0088] The updating module 305 includes:

[0089] The second repeated training unit is used to update the life decay model according to the newly added training samples of the target battery pack, and the update only changes the coefficients of the coefficient variable layer.

[0090] In combination with the embodiments described above, in the technical solution of the present application, when new training samples are added, the battery pack life decay model is updated by incremental learning, which can significantly improve the training efficiency of the model, save processor processing time, improve the model function, and improve the accuracy of the model in predicting and evaluating life decay changes.

[0091] Based on the method and apparatus for establishing a battery pack life attenuation model provided in the aforementioned embodiments, the present application also provides a device for establishing a battery pack life attenuation model.

[0092] The device includes a processor and a memory; the memory is used to store a computer program; the processor is used to execute the method for establishing a battery pack life attenuation model as described in the method embodiment according to the computer program.

[0093] An embodiment of the present application also provides a computer-readable storage medium for storing a computer program, and when the computer program is executed by a processor, the method for establishing a battery pack life attenuation model as described in the method embodiment is executed.

[0094] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for establishing a battery pack life attenuation model. It is characterized in that include: Establishing a life decay basic model for a target battery pack based on usage data and life decay data of a reference battery pack; the reference battery pack is a battery pack on other electric vehicles that were launched before the electric vehicle to which the target battery pack belongs; Obtaining usage data of the target battery pack; Performing transfer learning according to the usage data of the target battery pack and the usage data of the reference battery pack to obtain a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack; Obtaining a life decay model of the target battery pack according to the mapping relationship and the life decay basic model; When the training samples of the target battery pack increase, the life attenuation model of the target battery pack is updated through an incremental learning mode; The training samples include: usage data of a target battery pack and life attenuation data corresponding to the usage data; The life decay model of the target battery pack is established by using a support vector machine; the life decay model of the target battery pack is updated by using an incremental learning mode, including: Using the support vector machine to learn the first batch of training samples to obtain a first batch of support vectors; the first batch of support vectors are training samples selected by the support vector machine from the first batch of training samples; The support vector machine is trained again using a second batch of training samples and the first batch of support vectors to update the life decay model of the target battery pack; the second batch of training samples are training samples of the target battery pack that are newly added relative to the first batch of training samples.

2. The method according to claim 1, It is characterized in that The life decay model includes a fixed coefficient layer and a variable coefficient layer; The updating of the life attenuation model of the target battery pack by the incremental learning mode includes: The life decay model is updated according to the newly added training samples of the target battery pack, and the update only changes the coefficients of the coefficient variable layer.

3. The method according to any one of claims 1 to 2, It is characterized in that The reference battery pack is a battery pack of the same type as the target battery pack, or the reference battery pack is a battery pack of a similar type as the target battery pack.

4. A device for establishing a battery pack life attenuation model, It is characterized in that include: A basic model building module, used to build a life decay basic model of a target battery pack based on usage data and life decay data of a reference battery pack; the reference battery pack is a battery pack on other electric vehicles that were launched before the electric vehicle to which the target battery pack belongs; A target battery pack data acquisition module, used to obtain usage data of the target battery pack; a transfer learning module, configured to perform transfer learning according to the usage data of the target battery pack and the usage data of the reference battery pack, and obtain a mapping relationship between the usage data of the target battery pack and the usage data of the reference battery pack; A life decay model establishment module, used to obtain the life decay model of the target battery pack according to the mapping relationship and the life decay basic model; An updating module, configured to update the life attenuation model of the target battery pack through an incremental learning mode when the training samples of the target battery pack increase; The training samples include: usage data of a target battery pack and life attenuation data corresponding to the usage data; The life attenuation model of the target battery pack is established by using a support vector machine; the update module includes: A first acquisition unit is used to learn the first batch of training samples using the support vector machine to obtain a first batch of support vectors; the first batch of support vectors are training samples selected by the support vector machine from the first batch of training samples; The first repeated training unit is used to re-train the support vector machine using a second batch of training samples and the first batch of support vectors to update the life decay model of the target battery pack; the second batch of training samples are training samples of the target battery pack that are newly added relative to the first batch of training samples.

5. A device for establishing a battery pack life attenuation model, It is characterized in that It comprises a processor and a memory; the memory is used to store a computer program; the processor is used to execute the method for establishing a battery pack life attenuation model as described in any one of claims 1 to 3 according to the computer program.

6. A computer-readable storage medium, It is characterized in that Used to store a computer program, which, when executed by a processor, executes the method for establishing a battery pack life attenuation model as described in any one of claims 1 to 3.

Images