A method and device for quantifying the impact of different factors on battery health

Abstract

The present application discloses a method and apparatus for quantifying the impact of different factors on battery health, wherein the method includes: using the reduction in battery health within a first preset time period as a data label to label the statistical indicators of battery data within the first preset time period; the battery data includes at least two different factors; constructing a data set based on the labeled data; using the data set to train an attention model; using the trained attention model to quantify the impact of different factors in the battery data on battery health, and obtaining a quantified result. By using a specific data set to train a specific model, the model can quantify the impact of different factors on battery health, thereby quantifying the factors affecting the battery health (SOH).

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a method and device for quantifying the impact of different factors on battery health. Background Art

[0002] With the advent of the new energy era, driven by energy conservation, emission reduction, and environmental protection, batteries, as an important way to use new energy, have received increasing attention. Battery performance will greatly affect battery performance. Battery performance can be expressed by battery health (SOH). SOH is the ratio of the battery's current capacity to its rated capacity, reflecting the battery's current performance status.

[0003] There are many factors that affect the battery health SOH, and different battery users have different usage habits. Many factors such as the battery's operating temperature, current, and voltage will have a certain impact on the battery health SOH, so it is currently difficult to quantify the factors that affect the battery health SOH.

[0004] Therefore, how to quantify the factors that affect the battery health SOH has become a technical problem that needs to be solved urgently in this field. Summary of the Invention

[0005] Based on the above problems, the present application provides a method and device for quantifying the impact of different factors on battery health, so as to quantify the factors affecting the battery health SOH.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] In a first aspect, embodiments of the present application provide a method for quantifying the impact of different factors on battery health, including:

[0008] Using the reduction in battery health within a first preset time period as a data label, a statistical indicator of battery data within the first preset time period is labeled; the battery data includes at least two different factors;

[0009] Build a dataset based on the labeled data;

[0010] Training an attention model using the dataset;

[0011] The trained attention model is used to quantify the impact of different factors in the battery data on battery health and obtain quantitative results.

[0012] Optionally, a set of battery data includes:

[0013] at least two of the following: battery usage level, battery temperature when in use, battery output voltage, battery output current, battery non-use time, battery temperature when non-use, battery state of charge (SOC), and battery state of charge (SOC) when at rest; the battery usage level including vehicle mileage when the battery is used as a power source for a vehicle, mobile device usage time when the battery is used as a power source for a mobile device, or aircraft flight distance when the battery is used as a power source for an aircraft;

[0014] The statistical indicators include:

[0015] At least one of the maximum, minimum, mean, quantile, and standard deviation.

[0016] Optionally, the reduction in battery health within the first preset time period includes:

[0017] The difference between the average battery health within the second preset time period and the average battery health within the third preset time period; the starting point of the second preset time period is the starting point of the first preset time period, and the end point of the second preset time period is within the first preset time period; the end point of the third preset time period is the end point of the first preset time period, and the starting point of the third preset time period is within the first preset time period.

[0018] Optionally, the training of the attention model using the dataset includes:

[0019] Using the data set to train the attention model's ability to assign weights to different factors of the battery data;

[0020] The trained attention model is used to quantify the impact of different factors in the battery data on the battery health, and the quantitative results obtained include:

[0021] According to the result of the attention model assigning weights to different factors of the battery data, the impact of different battery data on battery health is quantified to obtain a quantified result.

[0022] Optionally, the trained attention model is used to quantify the impact of different factors in the battery data on the battery health, to obtain a quantified result, including:

[0023] Use the trained attention model to quantify the impact of different factors on battery health in a set of battery data, and obtain a single quantitative result;

[0024] A quantitative result is obtained according to multiple single quantitative results.

[0025] In a second aspect, an embodiment of the present application provides a device for quantifying the impact of different factors on battery health, including:

[0026] a labeling module, configured to label statistical indicators of battery data within a first preset time period using a reduction in battery health within the first preset time period as a data label; the battery data includes at least two different factors;

[0027] Dataset construction module, used to construct a dataset based on the labeled data;

[0028] A training module, configured to train an attention model using the dataset;

[0029] The quantification module is used to quantify the impact of different factors in the battery data on battery health using the trained attention model to obtain quantitative results.

[0030] Optionally, a set of battery data includes:

[0031] at least two of the following: battery usage level, battery temperature when in use, battery output voltage, battery output current, battery non-use time, battery temperature when non-use, battery state of charge (SOC), and battery state of charge (SOC) when at rest; the battery usage level including vehicle mileage when the battery is used as a power source for a vehicle, mobile device usage time when the battery is used as a power source for a mobile device, or aircraft flight distance when the battery is used as a power source for an aircraft;

[0032] The statistical indicators include:

[0033] At least one of the maximum, minimum, mean, quantile, and standard deviation.

[0034] Optionally, the reduction in battery health within the first preset time period includes:

[0035] The difference between the average battery health within the second preset time period and the average battery health within the third preset time period; the starting point of the second preset time period is the starting point of the first preset time period, and the end point of the second preset time period is within the first preset time period; the end point of the third preset time period is the end point of the first preset time period, and the starting point of the third preset time period is within the first preset time period.

[0036] Optionally, the training module includes:

[0037] an allocation capability training module, configured to train the ability of the attention model to allocate weights of different factors of the battery data using the data set;

[0038] The quantization module includes:

[0039] The distribution result quantification module is used to quantify the impact of different battery data on battery health according to the distribution result of the weights of different factors of the battery data by the attention model, and obtain a quantified result.

[0040] Optionally, the quantization module includes:

[0041] A single quantification module is used to quantify the impact of different factors on battery health in a set of battery data using the trained attention model, obtaining a single quantification result;

[0042] The synthesis module is used for obtaining a quantitative result based on multiple single quantitative results.

[0043] Compared with the existing technology, this application has the following beneficial effects:

[0044] The present application provides a method and apparatus for quantifying the impact of different factors on battery health. In the present application, the statistical indicators of battery data within a first preset time period are labeled by using the reduction in battery health within the first preset time period as a data label; the battery data includes at least two different factors; a data set is constructed based on the labeled data; the attention model is trained using the data set; and the trained attention model is used to quantify the impact of different factors in the battery data on battery health to obtain a quantitative result. By using a specific data set to train a specific model, the model can quantify the impact of different factors on battery health, thereby being able to quantify the factors affecting battery health SOH. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative labor.

[0046] Figure 1 A schematic flow chart of a method for quantifying the impact of different factors on battery health provided in an embodiment of the present application;

[0047] Figure 2 A flowchart of a method for obtaining a reduction in battery health within a first preset time period provided in an embodiment of the present application;

[0048] Figure 3 A schematic diagram of the structure of a device for quantifying the impact of different factors on battery health provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] As described above, there are many factors that affect the battery health SOH, so it is currently difficult to quantify the factors that affect the battery health SOH.

[0050] After research, the inventors have invented a method and device for quantifying the impact of different factors on the battery health, so as to quantify the factors affecting the battery health SOH.

[0051] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0052] Method Example

[0053] See also Figure 1 , which is a flow chart of a method for quantifying the impact of different factors on battery health provided by an embodiment of the present application, including the following steps:

[0054] S101 , using the reduction extent of battery health within a first preset time period as a data label to mark statistical indicators of battery data within the first preset time period; the battery data includes at least two different factors.

[0055] It should be noted that the first preset time period can be any time period with a fixed duration, and the battery data can be at least two of the battery usage level, the battery temperature when in use, the battery output voltage, the battery output current, the battery non-use time, the battery temperature when not in use, the battery state of charge SOC (state of charge) and the battery state of charge SOC when at rest; the battery usage level includes the vehicle mileage when the battery is used as a power source for the vehicle, the mobile device usage time when the battery is used as a power source for a mobile device, or the flight distance of the aircraft when the battery is used as a power source for an aircraft, that is, the battery data can be different when the battery is used for different purposes; the statistical indicator can be at least one of the maximum value, minimum value, mean, quantile and standard deviation, and the quantile can be the median, i.e., quantile, quartile, percentile, etc. It is understandable that in step S101, a set of labeled data may be the battery usage level, the maximum value of the battery temperature when in use, the minimum value of the battery temperature when in use, the mean value of the battery temperature when in use, the quantile of the battery temperature when in use, the standard deviation of the battery temperature when in use, the maximum value of the battery output voltage, the minimum value of the battery output voltage, the mean value of the battery output voltage, the quantile of the battery output voltage, the standard deviation of the battery output voltage, the maximum value of the battery output current, the minimum value of the battery output current, the mean value of the battery output current, the quantile of the battery output current, the standard deviation of the battery output current, the battery non-use time ... At least two different factors are selected from the group consisting of a maximum value of the battery temperature, a minimum value of the battery temperature when not in use, a mean value of the battery temperature when not in use, a quantile of the battery temperature when not in use, a standard deviation of the battery temperature when not in use, a maximum value of the battery state of charge SOC, a minimum value of the battery state of charge SOC, a mean value of the battery state of charge SOC, a quantile of the battery state of charge SOC, a standard deviation of the battery state of charge SOC, a maximum value of the battery state of charge SOC when at rest, a minimum value of the battery state of charge SOC when at rest, a quantile of the battery state of charge SOC when at rest, and a standard deviation of the battery state of charge SOC when at rest. The battery state of charge SOC when at rest may include the battery state of charge SOC when not in use.

[0056] In the embodiment provided in the present application, as an example, statistical indicators of battery data serving as a power source for a vehicle within three months can be counted as a feature space, and the reduction in battery health SOH within three months can be used as a data label to mark the corresponding data in the feature space.

[0057] S102: construct a data set based on the labeled data.

[0058] Specifically, in the example of step S101, the statistical indicators of the battery data serving as the power source of the vehicle within every three months are labeled with the corresponding reduction in the battery health SOH within three months as data labels. The labeled statistical indicators of the battery data serving as the power source of the vehicle within every three months can be used as a set of labeled data. Since there are currently multiple sets of labeled data, these labeled data can be used to construct a data set.

[0059] S103: Use the data set to train the attention model.

[0060] In the embodiment provided in this application, as an example, the data set can be used to train the attention model's ability to assign weights to different factors of the battery data;

[0061] It can be understood that what the above example means is that the data set can be used to train the attention model's ability to assign weights, and the weights are the weights of different factors of the battery data. That is, the attention model can assign weights to different factors contained in a set of battery data. By using the data set constructed in step S102 to train the attention model's ability to assign weights to different factors of the battery data, the attention model's ability to assign weights to different factors of the battery data can be improved, so that the attention model can more accurately assign weights to different factors of the battery data.

[0062] It should be noted that, because the embodiment provided in this application mainly uses tabular data for regression operations, an attention model structure specifically for tabular data can be selected. This structure is a serialized multi-step processing structure. The model has multiple modules. The attention module and the modules before the attention module process data in the following ways:

[0063] The data is separated by the feature processing module to be used for the Mask calculation of the attention module. The Mask is then used to perform feature screening with the input data processed by batch normalization (BN, BatchNorm). The module that has undergone feature screening will pass through the feature processing module f i It is divided into two parts d[i] and a[i], where d[i] is used to calculate the output of the current step and the input information of the remaining steps, and a[i] is used to calculate the attention module Mask. The specific formula is as follows:

[0064] [d[i],a[i]]=f i (Mask[i]·f)

[0065] The entire model is implemented in a cyclical manner. Except for the slightly different initial data splitting method, the calculation method of each other decision step can be the same, so it can be regarded as a large module.

[0066] The features after attention processing by Mask will continue to enter the next feature processing module and then split into two parts. The number of cycles is determined by the number of steps. The number of steps is a hyperparameter. It can be used in conjunction with the hyperparameter tuning algorithm or a common number of steps can be used. The common number of steps is generally 3.

[0067] In the embodiment provided in this application, the attention module used to calculate Mask[i] in each step is calculated as follows:

[0068] Mask[i]=sparse max(P[i-1]·h i (a[i-1]))

[0069] Mask[i] represents the value of the mask obtained by the attention module in the i-th decision step. Mask[i] can be used to calculate the attention level X corresponding to the vehicle data. a is the feature separated from the feature processing module and used for attention calculation. hi() represents BN+FC, which is a batch normalization (BN, BatchNorm) and fully connected (FC) operation. Prior scales scaling factor, the scaling factor of the i-th decision step is determined according to the mask of the previous i decision steps. If a feature has been selected many times in the previous steps, then the probability of the feature being selected should be reduced. The role of the scaling factor is to reduce the weight of this type of feature. γ is the degree of freedom parameter. When γ = 1, the feature is forced to appear only in one step. As the value of γ increases, this constraint is relaxed, and the number of feature occurrences increases. At the same time, the masked features are further split into new attention features a for the next attention calculation. Sparsemax is a sparser version of softmax. While softmax distributes the total attention of 1 to determine the attention contribution of each feature, sparsemax performs a sparser operation, assigning even the smallest values close to 0 to 0, enabling more robust decisions than softmax.

[0070] In the embodiment of the present application, the attention mechanism model can be used to obtain the attention corresponding to different factors of various battery data. As an example, the specific method for obtaining the attention is as follows:

[0071] When obtaining the attention corresponding to multiple factors, the input data is fed into the trained model. The model calculates mask[i] for each decision step. After obtaining multiple masks, they can be superimposed to obtain the final attention of the vehicle data. The number of multiple masks is determined by the number of decision steps, usually 3. The specific formula for superimposing masks to obtain the attention as an interpretable output is as follows:

[0072]

[0073] Where step_important[n] is the importance of step n, that is, the weight of the mask obtained at that step. n_step is the number of decision steps. Finally, the final superimposed mask is used to obtain the attention score. This means that by accumulating local information, global information is obtained, thereby obtaining the overall feature importance.

[0074] It should be noted that in the embodiments provided in this application, the attention model can automatically give high weights to important factors and low weights to unimportant factors.

[0075] S104: Use the trained attention model to quantify the impact of different factors in the battery data on the battery health and obtain a quantitative result.

[0076] Specifically, the influence of different battery data on battery health can be quantified based on the result of allocating weights of different factors of the battery data by the attention model to obtain a quantitative result.

[0077] It should be noted that, in step S104, the different factors in the battery data include multiple factors in the battery data in step S101. The quantization result may be the weight quantization value given by the attention model to the different factors.

[0078] In the embodiments provided herein, a trained attention model can be used to quantify the effects of different factors in a set of battery data on battery health, obtaining a single quantified result; and a quantified result can be obtained based on multiple single quantified results. After the attention model is trained, the attention model will assign certain weights to the factors in multiple sets of battery data, and the final quantified result can be obtained by combining the weights of the different factors in the multiple sets of battery data; the multiple sets of battery data can be the battery data in the data set constructed in step S102, or can be battery data obtained by other means.

[0079] An embodiment of the present application provides a method for quantifying the impact of different factors on battery health. The method uses the magnitude of the decrease in battery health within a first preset time period as a data label to annotate statistical indicators of battery data within the first preset time period. The battery data includes at least two different factors. A dataset is constructed based on the annotated data. An attention model is trained using the dataset. The trained attention model is used to quantify the impact of different factors in the battery data on battery health, obtaining a quantified result. By using a specific dataset to train a specific model, the model is able to quantify the impact of different factors on battery health, thereby quantifying the factors that affect the battery's state of health (SOH).

[0080] See also Figure 2 , which is a flow chart of a method for obtaining a reduction in battery health within a first preset time period provided by an embodiment of the present application, including the following steps:

[0081] S201, obtaining an average value of battery health within a second preset time period; the starting point of the second preset time period is the starting point of the first preset time period, and the ending point of the second preset time period is within the first preset time period.

[0082] It should be noted that the second preset time period is a part of the first preset time period, and the second preset time period may represent an initial period of time within the first preset time period.

[0083] Specifically, as an example, when the first preset time period is from 0:00 on July 1, 2020 to 0:00 on October 1, 2020, the duration of the second preset time period can be set to 10 days, that is, from 0:00 on July 1, 2020 to 0:00 on July 11, 2020. The average battery health SOH obtained in step S201 is the average battery health SOH of these 10 days.

[0084] It should be noted that in the embodiments provided herein, the battery health SOH at fixed time points each day within the second preset time period can be obtained, and the average value thereof can be used as the mean value of the battery health within the second preset time period. Alternatively, the battery health SOH value at each moment within the second preset time period can be continuously obtained and the mean value of the battery health within the second preset time period can be obtained by integration. Alternatively, the mean value of the battery health within the second preset time period can be obtained by other methods. This application does not limit the method for obtaining the mean value of the battery health within a certain time period.

[0085] S202, obtaining an average value of the battery health within a third preset time period; the end point of the third preset time period is the end point of the first preset time period, and the starting point of the third preset time period is within the first preset time period.

[0086] It should be noted that the third preset time period is a part of the first preset time period, and the third preset time period may represent a period of time ending in the first preset time period.

[0087] Specifically, as an example, when the first preset time period is from 0:00 on July 1, 2020 to 0:00 on October 1, 2020, the duration of the third preset time period can be set to 10 days, that is, from 0:00 on September 21, 2020 to 0:00 on October 1, 2020, and the average battery health SOH obtained in step S202 is the average battery health SOH of these 10 days.

[0088] In the embodiments provided herein, the battery health SOH values at fixed time points each day within the third preset time period can be obtained, and the average value can be used as the mean value of the battery health within the third preset time period. Alternatively, the battery health SOH values at each moment within the third preset time period can be continuously obtained and integrated to obtain the mean value of the battery health within the third preset time period. Alternatively, the mean value of the battery health within the third preset time period can be obtained by other methods. This application does not limit the method for obtaining the mean value of the battery health within a certain time period.

[0089] It should be noted that the starting point of the third preset time period is after the starting point of the second preset time period, and the end point of the second preset time period is before the end point of the third preset time period; the third preset time period and the second preset time period may or may not have an intersection.

[0090] The length of the third preset time period may be the same as or different from the length of the second preset time period. When the length of the third preset time period is the same as the length of the second preset time period, the extent of the reduction in battery health during the first preset time period can be more accurately determined.

[0091] S203: Taking the difference between the average value of the battery health in the second preset time period and the average value of the battery health in the third preset time period as the reduction range of the battery health in the first preset time period.

[0092] It should be noted that since the battery health shows a downward trend during normal use, in the embodiment provided in the present application, the average value of the battery health within the third preset time period is less than the average value of the battery health within the second preset time period, and the average value of the battery health within the second preset time period can represent the initial value of the battery health within the first preset time period, and the average value of the battery health within the third preset time period can represent the final value of the battery health within the first preset time period, so in step S703, the average value of the battery health within the second preset time period can be subtracted from the average value of the battery health within the third preset time period, and the obtained difference can be used as the reduction in battery health within the first preset time period.

[0093] The embodiment provided by this application uses the difference between the average battery health within the second preset time period and the average battery health within the third preset time period as the reduction in battery health within the first preset time period, thereby more accurately obtaining the reduction in battery health within a certain time period. In subsequent steps, the reduction in battery health within a certain time period is used as a data label to annotate the data, so that the training effect of the attention model trained using the annotated data as the data set is better, thereby enabling the trained attention model to better quantify the impact of different factors in the battery data on battery health.

[0094] Device embodiment

[0095] See also Figure 3 , this figure is a schematic diagram of the structure of a device for quantifying the impact of different factors on battery health provided in an embodiment of the present application, including: a labeling module 301, a data set construction module 302, a training module 303 and a quantization module 304.

[0096] The labeling module 301 is configured to label the statistical indicators of the battery data within a first preset time period using the reduction in battery health within the first preset time period as a data label; the battery data includes at least two different factors.

[0097] The data set construction module 302 is used to construct a data set based on the labeled data.

[0098] The training module 303 is used to train the attention model using the data set.

[0099] The quantification module 304 is used to quantify the impact of different factors in the battery data on the battery health using the trained attention model to obtain a quantified result.

[0100] Optionally, a set of battery data includes:

[0101] at least two of the following: battery usage level, battery temperature when in use, battery output voltage, battery output current, battery non-use time, battery temperature when non-use, battery state of charge (SOC), and battery state of charge (SOC) when at rest; the battery usage level including vehicle mileage when the battery is used as a power source for a vehicle, mobile device usage time when the battery is used as a power source for a mobile device, or aircraft flight distance when the battery is used as a power source for an aircraft;

[0102] The statistical indicators include:

[0103] At least one of the maximum, minimum, mean, quantile, and standard deviation.

[0104] Optionally, the reduction in battery health within the first preset time period includes:

[0105] The difference between the average battery health within the second preset time period and the average battery health within the third preset time period; the starting point of the second preset time period is the starting point of the first preset time period, and the end point of the second preset time period is within the first preset time period; the end point of the third preset time period is the end point of the first preset time period, and the starting point of the third preset time period is within the first preset time period.

[0106] Optionally, the training module 303 includes: an allocation capability training module for training the ability of the attention model to allocate weights of different factors of the battery data using the data set;

[0107] The quantization module 304 includes:

[0108] The distribution result quantification module is used to quantify the impact of different battery data on battery health according to the distribution result of the weights of different factors of the battery data by the attention model, and obtain a quantified result.

[0109] Optionally, the quantization module 304 includes:

[0110] A single quantification module is used to quantify the impact of different factors on battery health in a set of battery data using the trained attention model, obtaining a single quantification result;

[0111] The synthesis module is used for obtaining a quantitative result based on multiple single quantitative results.

[0112] An embodiment of the present application provides a device for quantifying the impact of different factors on battery health. The device uses the magnitude of the decrease in battery health within a first preset time period as a data label to annotate statistical indicators of battery data within the first preset time period. The battery data includes at least two different factors. A data set is constructed based on the annotated data. An attention model is trained using the data set. The trained attention model is used to quantify the impact of different factors in the battery data on battery health, obtaining a quantified result. By using a specific data set to train a specific model, the model is able to quantify the impact of different factors on battery health, thereby quantifying the factors that affect the battery's state of health (SOH).

[0113] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0114] The above is merely one specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for quantifying the impact of different factors on battery health, characterized in that: The method comprises: Using the reduction in battery health within a first preset time period as a data label, a statistical indicator of battery data within the first preset time period is labeled; the battery data includes at least two different factors; Build a dataset based on the labeled data; Training an attention model using the dataset; Use the trained attention model to quantify the impact of different factors in the battery data on battery health and obtain quantitative results; The training of the attention model using the data set includes: using the data set to train the attention model's ability to assign weights to different factors of the battery data; wherein the training method is as follows: For the data in the data set, the features of the data are processed by the feature processing module f i It is divided into two parts: d[i] and a[i]; wherein, d[i] is used to calculate the output of the current step and the input information of the remaining steps, and a[i] is used to calculate the attention module Mask, as described in the following formula: [d[i],a[i]]=f i Mask[i]f) The features after attention processing by Mask will continue to enter the next feature processing module and then split into two parts. The number of cycles is determined by the number of steps. The attention module used to calculate Mask[i] in each step is as described in the following formula: Mask[i]=sparsemax(P[i-1]·h i (a[i1])) Among them, Mask[i] represents the value of the mask obtained by the attention module in the i-th decision step, a represents the feature separated from the feature processing module, and h i (a[i-1]) represents the normalization and full connection operation of the i-1th feature separated from the feature processing module, P[i-1] is the Prior scales scaling factor of the i-1th decision step, sparsemax is the sparse version of softmax, and softmax distributes the total attention of 1 to obtain the attention ratio of each feature; After obtaining multiple masks, we superimpose them using the following formula to obtain the final attention of the vehicle data: Among them, step_importance[n] is the importance of the nth step, n_step is the number of decision steps, and mask[n] represents the nth mask value; The trained attention model is used to quantify the impact of different factors in the battery data on the battery health to obtain a quantified result, including: quantifying the impact of different battery data on the battery health according to the weight distribution result of different factors of the battery data by the attention model to obtain a quantified result.

2. The method according to claim 1, characterized in that A set of battery data includes: at least two of the following: battery usage level, battery temperature when in use, battery output voltage, battery output current, battery non-use time, battery temperature when non-use, battery state of charge (SOC), and battery state of charge (SOC) when at rest; the battery usage level including vehicle mileage when the battery is used as a power source for a vehicle, mobile device usage time when the battery is used as a power source for a mobile device, or aircraft flight distance when the battery is used as a power source for an aircraft; The statistical indicators include: At least one of the maximum, minimum, mean, quantile, and standard deviation.

3. The method according to claim 1, characterized in that The reduction in battery health within the first preset time period includes: The difference between the average battery health within the second preset time period and the average battery health within the third preset time period; the starting point of the second preset time period is the starting point of the first preset time period, and the end point of the second preset time period is within the first preset time period; the end point of the third preset time period is the end point of the first preset time period, and the starting point of the third preset time period is within the first preset time period.

4. The method according to claim 1, wherein The trained attention model is used to quantify the impact of different factors in the battery data on the battery health, and the quantitative results obtained include: Use the trained attention model to quantify the impact of different factors on battery health in a set of battery data, and obtain a single quantitative result; A quantitative result is obtained according to multiple single quantitative results.

5. A device for quantifying the impact of different factors on battery health, characterized in that: The device comprises: a labeling module, configured to label statistical indicators of battery data within a first preset time period using a reduction in battery health within the first preset time period as a data label; the battery data includes at least two different factors; Dataset construction module, used to construct a dataset based on the labeled data; A training module, configured to train an attention model using the dataset; The quantification module is used to quantify the impact of different factors in the battery data on battery health using the trained attention model to obtain quantitative results; The training module is specifically used to train the attention model's ability to assign weights to different factors of the battery data using the data set; the training method is as follows: For the data in the data set, the features of the data are processed by the feature processing module f i It is divided into two parts: d[i] and a[i]; wherein, d[i] is used to calculate the output of the current step and the input information of the remaining steps, and a[i] is used to calculate the attention module Mask, as described in the following formula: [d[i],a[i]]=f i (Mask[i]·f) The features after attention processing by Mask will continue to enter the next feature processing module and then split into two parts. The number of cycles is determined by the number of steps. The attention module used to calculate Mask[i] in each step is as described in the following formula: Mask[i]=sparsemax(P[i-1]·h i (a[i-1])) Among them, Mask[i] represents the value of the mask obtained by the attention module in the i-th decision step, a represents the feature separated from the feature processing module, and h i (a[i-1]) represents the normalization and full connection operation of the i-1th feature separated from the feature processing module, P[i-1] is the Prior scales scaling factor of the i-1th decision step, sparsemax is the sparse version of softmax, and softmax distributes the total attention of 1 to obtain the attention ratio of each feature; After obtaining multiple masks, we superimpose them using the following formula to obtain the final attention of the vehicle data: Among them, step_importance[n] is the importance of the nth step, n_step is the number of decision steps, and mask[n] represents the nth mask value; The quantification module is specifically used to quantify the impact of different battery data on battery health based on the distribution result of the weights of different factors of the battery data by the attention model, and obtain a quantification result.

6. The device according to claim 5, characterized in that A set of battery data includes: at least two of the following: battery usage level, battery temperature when in use, battery output voltage, battery output current, battery non-use time, battery temperature when non-use, battery state of charge (SOC), and battery state of charge (SOC) when at rest; the battery usage level including vehicle mileage when the battery is used as a power source for a vehicle, mobile device usage time when the battery is used as a power source for a mobile device, or aircraft flight distance when the battery is used as a power source for an aircraft; The statistical indicators include: At least one of the maximum, minimum, mean, quantile, and standard deviation.

7. The device according to claim 5, characterized in that The reduction in battery health within the first preset time period includes: The difference between the average battery health within the second preset time period and the average battery health within the third preset time period; the starting point of the second preset time period is the starting point of the first preset time period, and the end point of the second preset time period is within the first preset time period; the end point of the third preset time period is the end point of the first preset time period, and the starting point of the third preset time period is within the first preset time period.

8. The device according to claim 5, characterized in that The quantization module includes: A single quantification module is used to quantify the impact of different factors on battery health in a set of battery data using the trained attention model, obtaining a single quantification result; The synthesis module is used for obtaining a quantitative result based on multiple single quantitative results.

Images