Method for training SOH prediction model of battery and method for predicting SOH of battery

Abstract

The invention discloses a method for training a battery SOH prediction model and a method for predicting the SOH of a battery, and the method comprises the steps: obtaining a training sample set, the training sample set comprises a plurality of training samples, and each training sample comprises a historical feature value of a target physical aging feature of the battery and a corresponding tag SOH; sequentially inputting training samples of the training sample set into a to-be-trained model for iterative training, and determining a target loss value of the to-be-trained model in each training according to physical constraints associated with the target physical aging characteristics in the iterative training; and under the condition that the target loss value meets a preset iteration stop condition, training to obtain the battery SOH prediction model. Therefore, the precision of the battery SOH prediction model can be improved.

Description

Technical Field

[0001] The present application belongs to the field of vehicle technology, and in particular relates to a method for training a battery SOH prediction model and a method for predicting a battery SOH. Background Art

[0002] With the rapid development of vehicle technology and the increasing awareness of environmental protection, electric vehicles (EVs), which offer advantages such as zero emissions, low pollution, and high energy efficiency, are becoming increasingly popular. Batteries, as a crucial component of EVs, primarily provide energy for the EV's drive motors, enabling them to propel the EV forward.

[0003] Currently, in the battery management process of electric vehicles, trained machine learning models are often used to predict the battery's State of Health (SOH). However, machine learning models trained using traditional model training methods may not match the actual battery application scenarios, resulting in low prediction accuracy. Summary of the Invention

[0004] The embodiments of the present application provide a method for training a battery SOH prediction model and a method for predicting the SOH of a battery, which can improve the prediction accuracy of the trained machine learning model.

[0005] In a first aspect, an embodiment of the present application provides a method for training a battery SOH prediction model, the method comprising:

[0006] Acquire a training sample set, the training sample set including a plurality of training samples, each of the training samples including a historical characteristic value of a target physical aging characteristic of a battery and a corresponding label SOH;

[0007] Inputting the training samples of the training sample set into the model to be trained in sequence for iterative training, and determining the target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics during the iterative training;

[0008] When the target loss value meets the preset iteration stop condition, the battery SOH prediction model is trained.

[0009] In some embodiments, the target physical aging characteristics include:

[0010] The battery's operating characteristics of continuous constant current within the preset charging time;

[0011] The accumulated characteristic quantity of the battery throughout its life cycle.

[0012] In some embodiments, the physical constraint comprises a first physical constraint;

[0013] The step of determining a target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics during the iterative training includes:

[0014] When an Nth training sample is input into the model to be trained for training, in response to the first physical constraint, obtaining an Nth first predicted SOH pre-output by the model to be trained, where N is a positive integer;

[0015] If the Nth first predicted SOH exceeds a preset reference SOH range, adjusting the Nth first predicted SOH to be within the reference SOH range, where the reference SOH range is associated with the accumulated feature quantity in the Nth training sample;

[0016] According to the adjusted Nth first predicted SOH, the target loss value of the to-be-trained model in the Nth training is calculated.

[0017] In some embodiments, the physical constraint includes a second physical constraint;

[0018] The step of determining a target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics during the iterative training includes:

[0019] When the Mth training sample is input into the to-be-trained model for training, in response to the second physical constraint, obtaining a target number of training times in the M training times that satisfies a preset condition, wherein the preset condition is that a change trend of the second predicted SOH in two previous and subsequent training times matches a change trend of the accumulated feature value, where M is a positive integer;

[0020] According to the ratio of the target number of training times to M, the target loss value of the to-be-trained model in the Mth training is calculated.

[0021] In some embodiments, obtaining a target number of training sessions that meets a preset condition in the M training sessions includes:

[0022] Calculating a first difference in the second predicted SOH between the Mth training and the M-1th training, and calculating a second difference in the accumulated feature quantities in the training samples used in the Mth training and the M-1th training;

[0023] Determining, based on the first difference and the second difference, a matching result of a change trend of the second predicted SOH and a change trend of the accumulated feature value between the M-th training and the M-1-th training;

[0024] generating a target symbol if the matching result indicates that the change trend of the second predicted SOH matches the change trend of the accumulated feature amount;

[0025] The target number of training times is the number of target symbols generated in the M training times.

[0026] In some embodiments, the second predicted SOH is obtained by adjusting the first predicted SOH output by the to-be-trained model in the Mth training in response to the first physical constraint.

[0027] The step of calculating the target loss value of the model to be trained in the Mth training according to the ratio of the target number of training times to M includes:

[0028] Calculate the first loss value of the to-be-trained model in the Mth training round according to the second predicted SOH; and use the ratio of the target number of training times to M as the second loss value of the to-be-trained model in the Mth training round;

[0029] The first loss value and the second loss value are fused to obtain the target loss value.

[0030] In some embodiments, fusing the first loss value and the second loss value to obtain the target loss value includes:

[0031] If the training sample set includes K training samples, and the Mth sample is a training sample between the K / 2th training sample and the Kth training sample, the first loss value and the second loss value are fused to obtain the target loss value, where K is an integer greater than 1;

[0032] The step of determining a target loss value of the model to be trained in each training session according to the physical constraints associated with the target physical aging characteristics during the iterative training further includes:

[0033] If the Mth training sample is a training sample between the 1st training sample and the K / 2th training sample, the first loss value is determined as the target loss value.

[0034] In some embodiments, the first loss value and the second loss value are fused to obtain the target loss value, including:

[0035] The sum of the first loss value and a target product is determined as the target loss value, where the target product is the product of the second loss value and a target weight, and the target weight is associated with a sorting position of the Mth training sample in the K training samples.

[0036] In some implementations, before obtaining the training sample set, the method further includes:

[0037] Acquire multiple groups of historical characteristic data of the battery, each group of the historical characteristic data including historical characteristic values ​​of multiple physical aging characteristics;

[0038] determining, based on the multiple sets of historical characteristic data, a degree of correlation between the multiple physical aging characteristics;

[0039] selecting at least some of the physical aging features from the plurality of physical aging features as the target physical aging features according to the correlation between the plurality of physical aging features;

[0040] The training sample set is generated based on the historical feature values ​​of the target physical aging feature in the multiple sets of historical feature data.

[0041] In a second aspect, an embodiment of the present application further provides a method for predicting the SOH of a battery, the method comprising:

[0042] Obtaining the current characteristic value of the target physical aging characteristic of the battery under test;

[0043] Input the current characteristic value of the target physical aging characteristic into the battery SOH prediction model to obtain the predicted SOH of the battery to be tested,

[0044] The battery SOH prediction model is obtained by training the battery SOH prediction model according to any one of the methods for training the battery SOH prediction model in the first aspect.

[0045] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor and a memory storing computer program instructions;

[0046] When the processor executes the computer program instructions, it implements the method for training a battery SOH prediction model as described in any one of the first aspects or the method for predicting the SOH of a battery as described in the second aspect.

[0047] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implements the method for training a battery SOH prediction model as described in any one of the first aspects or the method for predicting the SOH of a battery as described in the second aspect.

[0048] In a fifth aspect, an embodiment of the present application further provides a vehicle comprising the electronic device as described in the third aspect.

[0049] In an embodiment of the present application, a training sample set is obtained, the training sample set includes multiple training samples, each of the training samples includes a historical characteristic value of the target physical aging characteristic of the battery and a corresponding label SOH; the training samples of the training sample set are sequentially input into the model to be trained for iterative training, and in the iterative training, the target loss value of the model to be trained in each training is determined according to the physical constraints associated with the target physical aging characteristic; when the target loss value meets the preset iteration stop condition, the battery SOH prediction model is trained. In this way, the loss value of the model to be trained during the training process can be affected based on the physical constraints associated with the target physical aging characteristic, so that the trained battery SOH prediction model is more suitable for the actual application scenario of the battery, thereby improving the accuracy of the battery SOH prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0051] Figure 1 1 is a flow chart of an embodiment of a method for training a battery SOH prediction model provided by the present application;

[0052] Figure 2 1 is a flow chart of an embodiment of a method for predicting the SOH of a battery provided by the present application;

[0053] Figure 3 1 is a schematic structural diagram of an embodiment of a device for training a battery SOH prediction model provided by the present application;

[0054] Figure 4 1 is a schematic structural diagram of an embodiment of a device for predicting SOH of a battery provided by the present application;

[0055] Figure 5 It is a structural diagram of an embodiment of the electronic device provided by this application. DETAILED DESCRIPTION

[0056] The features and exemplary embodiments of various aspects of the present application will be described in detail below. To make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present application by illustrating the examples of the present application.

[0057] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, the elements defined by the phrase "comprising..." do not exclude the presence of other identical elements in the process, method, article, or device comprising the elements.

[0058] Figure 1 The flowchart of the embodiment of the method for training the battery SOH prediction model provided by the present application is shown in FIG. The method for training the battery SOH prediction model provided by the present application can be applied to an electronic device, which can be an onboard terminal of a vehicle or a terminal device, server, etc. independent of the vehicle. Figure 1 As shown, the method for training the battery SOH prediction model includes but is not limited to the following steps S101 to S103.

[0059] Step S101: Acquire a training sample set, where the training sample set includes multiple training samples. Each training sample includes a historical feature value of a target physical aging feature of a battery and a corresponding label SOH.

[0060] In this step, the electronic device may obtain a training sample set including a plurality of training samples.

[0061] The training samples in the above training sample set can be pre-generated, and the process of generating each training sample can include: obtaining the historical characteristic value of the target physical aging characteristic of the battery, and obtaining the label SOH manually input for the historical characteristic value; forming a training sample including the historical characteristic value and the SOH label.

[0062] The target physical aging characteristics mentioned above may be any physical aging characteristics used to characterize the aging conditions that occur during the use of the battery.

[0063] In some embodiments, the target physical aging characteristics may include:

[0064] The battery's operating characteristics of continuous constant current within the preset charging time;

[0065] The accumulated characteristic quantity of the battery throughout its life cycle.

[0066] In this embodiment, since the target physical aging characteristics include the operating condition characteristics of the battery at a continuous constant current within the preset charging time and the cumulative characteristic quantities throughout the battery's life cycle, more comprehensive battery status information can be provided, further improving the accuracy of the trained battery SOH prediction model.

[0067] Among them, the above-mentioned operating condition characteristics may include at least one of the charging start SOC, charging end SOC, charging start temperature, charging end temperature, charging start voltage, charging end voltage, charging process current, charging process temperature variance, charging process voltage variance, charging process voltage curve entropy, charging process temperature curve entropy, charged amount, maximum temperature difference during charging process, and maximum pressure difference during charging process; the above-mentioned cumulative characteristic quantity may include the cumulative discharge amount.

[0068] The target physical aging characteristics may include all characteristics of the battery's operating condition characteristics at a continuous constant current within the preset charging time and the accumulated characteristic value.

[0069] For example, at a certain moment, the charging start SOC, charging end SOC, charging start temperature, charging end temperature, charging start voltage, charging end voltage, charging process current, charging process temperature variance, charging process voltage variance, charging process voltage curve entropy, charging process temperature curve entropy, charged capacity, maximum temperature difference during the charging process, and maximum voltage difference during the charging process at a point where the battery has a continuous constant current within a preset charging time before the moment can be collected, as well as the cumulative discharge capacity of the battery at that moment; and all the collected characteristic values ​​are used as the historical characteristic values ​​of the above-mentioned target physical aging characteristics.

[0070] Alternatively, the target physical aging characteristics may include at least part of the operating condition characteristics of the battery at a continuous constant current within the preset charging time and the cumulative characteristic value, and the at least part of the characteristics may be obtained by screening from all the operating condition characteristics manually or by a preset screening method.

[0071] In some implementations, before obtaining the training sample set, the method further includes:

[0072] Acquire multiple sets of historical characteristic data of the battery, each set of historical characteristic data including historical characteristic values ​​of multiple physical aging characteristics;

[0073] Determining the degree of correlation between multiple physical aging characteristics based on multiple sets of historical characteristic data;

[0074] screening at least some of the physical aging features as target physical aging features from the multiple physical aging features according to correlations between the multiple physical aging features;

[0075] A training sample set is generated based on historical feature values ​​of target physical aging features in multiple sets of historical feature data.

[0076] In this embodiment, at least some of the physical aging features can be screened as target physical aging features based on the degree of correlation between the multiple physical aging features, thereby ensuring the coupling relationship between the physical aging features in the target physical aging features, improving the rationality of the input parameters of the battery SOH prediction model, and further improving the accuracy of the trained battery SOH prediction model.

[0077] Acquiring multiple sets of historical characteristic data of the battery can involve collecting historical characteristic data at multiple different moments as the multiple sets of historical characteristic data. It should be noted that all historical characteristic data collected at each time can be used as data in the multiple sets of historical characteristic data; alternatively, after collecting a raw data set containing a certain amount of historical characteristic data, the data in the raw data set can be preprocessed to remove historical characteristic data that does not meet the requirements, and the historical characteristic data in the preprocessed raw data set can be used as the multiple sets of historical characteristic data.

[0078] The above-mentioned preprocessing of the data in the original data set may include at least one of the following:

[0079] Based on the data characteristics of the battery, invalid temperature, voltage, and current data are screened out, especially data that exceeds the pre-defined valid range;

[0080] Based on the real-time response of current and voltage and data synchronization, abnormal data is screened out during the process.

[0081] Determining the correlation between the multiple physical aging features based on the multiple sets of historical feature data may be performed by analyzing the correlation between the multiple physical aging features using a preset correlation analysis strategy. The correlation analysis strategy may include at least one of multiple correlation analysis strategies, such as Pearson Correlation Analysis (PCA), Gray Correlation Analysis (GCA), and Maximum Information Coefficient (MIC).

[0082] In some embodiments, before the above-mentioned analysis of the correlation degree between multiple physical aging characteristics using a preset correlation analysis strategy, it may also include: determining a target correlation analysis strategy among multiple preset correlation analysis strategies; the above-mentioned analysis of the correlation degree between multiple physical aging characteristics using a preset correlation analysis strategy may include: the above-mentioned analysis of the correlation degree between multiple physical aging characteristics using a correlation analysis strategy.

[0083] The target correlation analysis strategy is determined among the plurality of preset correlation analysis strategies. The user may select at least one correlation analysis strategy from the plurality of correlation analysis strategies based on experience, and in response to the selection operation, the selected at least one correlation analysis strategy may be used as the target correlation analysis strategy.

[0084] Alternatively, the target correlation analysis strategy may be determined from among the plurality of preset correlation analysis strategies, and the electronic device may automatically select the target correlation analysis strategy based on the plurality of sets of historical feature data.

[0085] For example, the electronic device may judge the input data set (i.e., multiple sets of historical feature data) to determine whether there are missing values ​​in the data set (or, the missing values ​​are greater than a certain threshold). If so, the grayscale correlation analysis strategy is preferentially selected; if there are no missing values, the linear correlation analysis strategy between features is performed; if the proportion of data in the data set that conforms to the linear relationship exceeds a preset threshold (i.e., most of them conform to the linear relationship), the Pearson correlation analysis is selected; or, if the proportion of data in the data set that conforms to the linear relationship is lower than the preset threshold (i.e., there are highly obvious nonlinear correlation features), the maximum information coefficient is selected.

[0086] The correlations among the multiple physical aging characteristics may be correlations between each physical aging characteristic and each other physical aging characteristic. Selecting at least some of the multiple physical aging characteristics as target physical aging characteristics based on the correlations among the multiple physical aging characteristics may include selecting physical aging characteristics having correlations greater than a preset correlation threshold as target physical aging characteristics.

[0087] For example, assuming that the preset correlation threshold is 0.8, if it is determined that the correlation between the charged power and the maximum temperature difference during the charging process is greater than or equal to 0.8, the charged power and the maximum temperature difference during the charging process are determined as features in the target physical aging features.

[0088] Step S102: input the training samples of the training sample set into the model to be trained in sequence for iterative training, and determine the loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics during the iterative training.

[0089] In this step, the above-mentioned physical constraints can be pre-configured to be related to the target physical aging characteristics of the battery, and to generate constraints on the loss value during the training process of the model to be trained.

[0090] In some embodiments, the physical constraint includes a first physical constraint; and determining the target loss value of the model to be trained in each training session according to the physical constraint associated with the target physical aging characteristic during the iterative training includes:

[0091] When the Nth training sample is input into the to-be-trained model for training, in response to the first physical constraint, obtaining the Nth first predicted SOH pre-output by the to-be-trained model, where N is a positive integer;

[0092] If the Nth first predicted SOH exceeds a preset reference SOH range, the Nth first predicted SOH is adjusted to be within the reference SOH range, where the reference SOH range is associated with the accumulated feature quantity in the Nth training sample;

[0093] According to the adjusted Nth first predicted SOH, the target loss value of the model to be trained in the Nth training is calculated.

[0094] In this embodiment, the electronic device can, based on the first physical constraint, limit the first predicted SOH to the reference SOH range when the first predicted SOH output by the model to be trained exceeds the preset reference SOH range during the training process, thereby realizing the constraint on the target loss value based on the reference SOH range, thereby improving the accuracy of the trained battery SOH prediction model.

[0095] The Nth training sample may be a training sample input in any training session during the training process.

[0096] The above-mentioned electronic device may be configured with preset SOH ranges corresponding to different cumulative feature quantities, and when the Nth training sample is input into the model to be trained, the electronic device may determine the preset SOH range corresponding to the cumulative feature quantity as the above-mentioned reference SOH range based on the cumulative feature quantity in the historical feature value of the Nth training sample.

[0097] For example, the electronic device may be pre-configured with the SOH upper limit and SOH lower limit table shown in Table 1 below:

[0098] Table 1 SOH upper limit and SOH lower limit table

[0099]

[0100] It can be seen that in Table 1, the preset SOH range corresponding to different cumulative discharge amounts is from the lower limit SOH (i.e., SOH_B) to the upper limit SOH (i.e., SOH_A), and the electronic device can select the corresponding range as the reference SOH range based on the cumulative discharge amount in the input training sample.

[0101] For example, when the cumulative discharge amount in the Nth training sample is within the cumulative discharge amount range of [0, 4500), the corresponding reference SOH may be [0.9, 1]; when the cumulative discharge amount in the Nth training sample is within the cumulative discharge amount range of [4500, 15000), the corresponding reference SOH may be [0.891, 0.99]; when the cumulative discharge amount in the Nth training sample is within the cumulative discharge amount range of [15000, 30000), the corresponding reference SOH may be [0.873, 0.97], ..., and so on.

[0102] If the Nth first predicted SOH exceeds the preset reference SOH range, the Nth first predicted SOH is adjusted to be within the reference SOH range. The first predicted SOH can be adjusted to any SOH within the reference SOH range, for example, to an intermediate SOH.

[0103] In some embodiments, the above-mentioned adjustment of the Nth first predicted SOH to be within the reference SOH range can be that when the first predicted SOH is lower than the lower limit SOH of the reference SOH range, the first predicted SOH is adjusted to the lower limit SOH; and when the first predicted SOH is higher than the upper limit SOH of the reference SOH range, the first predicted SOH is adjusted to the upper limit SOH.

[0104] For example, when the cumulative discharge amount in the Nth training sample is within [15000, 30000) and the reference SOH is [0.873, 0.97], if the Nth first predicted SOH is lower than 0.873, the Nth first predicted SOH is set to 0.873; if the Nth first predicted SOH is higher than 0.97, the Nth first predicted SOH is set to 0.97.

[0105] The target loss value of the model to be trained in the Nth training round is calculated based on the adjusted Nth first predicted SOH. The target loss value can be calculated using any loss function. For example, the mean absolute error or mean square error of the adjusted Nth first predicted SOH can be calculated.

[0106] In some embodiments, the above calculation of the target loss value of the to-be-trained model in the Nth training according to the adjusted Nth first predicted SOH includes:

[0107] Calculate the average root mean square error between the Nth first predicted SOH and the corresponding label SOH;

[0108] According to the average root mean square error, the target loss value of the model to be trained in the Nth training is determined.

[0109] In this embodiment, by taking the average root mean square error between the Nth first predicted SOH and the corresponding label SOH as the target loss value, the accuracy of the target loss value can be improved, thereby further improving the accuracy of the battery SOH prediction model.

[0110] The target loss value of the model to be trained in the Nth training cycle is determined based on the average root mean square error. The average root mean square error can be directly used as the target loss function.

[0111] In some embodiments, the physical constraint includes a second physical constraint; determining a target loss value of the model to be trained in each training session according to the physical constraint associated with the target physical aging characteristic during iterative training includes:

[0112] When the Mth training sample is input into the to-be-trained model for training, in response to the second physical constraint, a target number of training times satisfying a preset condition in the M training times is obtained, where the preset condition is that a change trend of the second predicted SOH in two previous and subsequent training times matches a change trend of the accumulated feature quantity, and M is a positive integer;

[0113] According to the ratio of the target number of training times to M, the target loss value of the model to be trained in the Mth training is calculated.

[0114] In this embodiment, in the Mth training, the target loss value can be calculated based on the proportion of the target training times that meet the preset conditions in the M training times, thereby realizing the constraint on the target loss value based on the change trend of the second predicted SOH in the two previous and subsequent trainings and the change trend of the cumulative characteristic quantity, thereby improving the accuracy of the battery SOH prediction model obtained through training.

[0115] The above-mentioned acquisition of the target number of trainings that meet the preset conditions in M ​​trainings can be that a counter is set in the electronic device, and in the M trainings, if the change trend of the second predicted SOH in the current training and the previous training matches the change trend of the accumulated feature quantity, the count value of the counter is increased by 1, otherwise it is not counted; when the change trend matching result between the Mth training and the M-1th training is determined, the count value of the counter is determined as the target number of trainings.

[0116] The above-mentioned electronic device determines whether the changing trend of the second predicted SOH in the two previous and subsequent trainings matches the changing trend of the cumulative feature quantity. It can be by calculating the first difference of the second predicted SOH between the Mth training and the M-1th training, and calculating the second difference between the cumulative feature quantities in the training samples used for the Mth training and the M-1th training, and then determining the matching results of the changing trend of the second predicted SOH and the changing trend of the cumulative feature quantity between the Mth training and the M-1th training based on the first difference and the second difference.

[0117] For example, in the case where the above-mentioned cumulative feature quantity includes the cumulative discharge amount, if the value of the first difference is positive and the value of the second difference is positive, it is determined that the second predicted SOH in the two previous and subsequent trainings increases as the cumulative discharge amount increases, and in this case, the matching result is determined to indicate that the change trend of the second predicted SOH matches the change trend of the cumulative feature quantity; if the value of the first difference is negative and the value of the second difference is positive, it is determined that the second predicted SOH in the two previous and subsequent trainings decreases as the cumulative discharge amount increases, and in this case, the matching result is determined to indicate that the change trend of the second predicted SOH does not match the change trend of the cumulative feature quantity;

[0118] Alternatively, in the case where the above-mentioned cumulative characteristic quantity includes the remaining discharge capacity, and the remaining discharge capacity is the difference between the rated discharge capacity of the battery and the cumulative discharge capacity, if the value of the first difference is positive and the value of the second difference is negative, then it is determined that the second predicted SOH in the two previous and subsequent trainings increases as the remaining discharge capacity decreases, and at this time, the matching result is determined to indicate that the changing trend of the second predicted SOH matches the changing trend of the cumulative characteristic quantity; if the value of the first difference is negative and the value of the second difference is negative, then it is determined that the second predicted SOH in the two previous and subsequent trainings decreases as the remaining discharge capacity decreases, and at this time, the matching result is determined to indicate that the changing trend of the second predicted SOH does not match the changing trend of the cumulative characteristic quantity.

[0119] Alternatively, in some embodiments, obtaining the target number of training sessions that meets a preset condition in the M training sessions may include:

[0120] Calculating a first difference in the second predicted SOH between the Mth training and the M-1th training, and calculating a second difference in the accumulated feature quantities in the training samples used in the Mth training and the M-1th training;

[0121] Determine, based on the first difference and the second difference, a matching result of a change trend of the second predicted SOH and a change trend of the accumulated feature quantity between the Mth training and the M-1th training;

[0122] generating a target symbol when the matching result indicates that the change trend of the second predicted SOH matches the change trend of the accumulated feature amount;

[0123] The target number of training times is the number of target symbols generated in M ​​training times.

[0124] In this embodiment, the electronic device can generate a target symbol based on the matching results of the changing trend of the second predicted SOH and the changing trend of the cumulative feature quantity between the Mth training and the M-1th training, thereby realizing the symbolization of each matching result, making it more convenient to determine the target number of training times.

[0125] The target symbol can be any numerical value, character string or icon.

[0126] In addition, when the above matching result indicates that the change trend of the second predicted SOH does not match the change trend of the accumulated feature amount, the electronic device can generate other symbols different from the target symbol.

[0127] For example, when the above-mentioned cumulative characteristic quantity includes the cumulative discharge amount, if it is determined that the second predicted SOH in the two previous and subsequent trainings increases with the increasing cumulative discharge amount, the value is 1 (i.e., the target symbol); if it is determined that the second predicted SOH in the two previous and subsequent trainings decreases with the increasing cumulative discharge amount, the value is 0 (i.e., other symbols).

[0128] The target loss value of the model to be trained in the Mth training is calculated based on the ratio of the target number of training times to M. The ratio of the target number of training times to M can be directly used as the target loss value.

[0129] It should be noted that the second predicted SOH mentioned above may be the predicted SOH directly output by the to-be-trained model in the Mth training.

[0130] In some embodiments, the second predicted SOH is obtained by adjusting the first predicted SOH output by the model to be trained in the Mth training round in response to the first physical constraint; and the target loss value of the model to be trained in the Mth training round is calculated based on the ratio of the target number of training times to M, including:

[0131] Calculate the first loss value of the model to be trained in the Mth training round based on the second predicted SOH; and use the ratio of the target number of training rounds to M as the second loss value of the model to be trained in the Mth training round.

[0132] The first loss value and the second loss value are fused to obtain the target loss value.

[0133] In this embodiment, the first loss value generated based on the first physical constraint and the second loss value generated based on the second physical constraint can be fused to obtain the target loss value, thereby combining the first physical constraint and the second physical constraint to influence the model training process and further improve the accuracy of the battery SOH prediction model.

[0134] It should be noted that the above-mentioned fusion of the first loss value generated based on the first physical constraint and the second loss value generated based on the second physical constraint to obtain the target loss value can be performed in any training during the model training process.

[0135] Alternatively, in some embodiments, fusing the first loss value and the second loss value to obtain the target loss value may include:

[0136] If the training sample set includes K training samples, and the Mth sample is a training sample between the K / 2th training sample and the Kth training sample, the first loss value and the second loss value are fused to obtain the target loss value, where K is an integer greater than 1.

[0137] The above-mentioned determination of the target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics in the iterative training may also include:

[0138] If the Mth training sample is a training sample between the 1st training sample and the K / 2th training sample, the first loss value is determined as the target loss value.

[0139] In this embodiment, the electronic device can use the first loss value generated based on the first physical constraint as the target loss value in the first half of the training; and in the second half of the training, the first loss value generated based on the first physical constraint and the second loss value generated based on the second physical constraint are integrated to obtain the target loss value, thereby realizing the phased use of different physical constraints to influence model training, thereby improving rationality and further improving the accuracy of the battery SOH prediction model.

[0140] The first loss value and the second loss value are fused to obtain the target loss value, and the sum of the first loss value and the second loss value can be directly used as the target loss value.

[0141] In some embodiments, the first loss value and the second loss value are fused to obtain a target loss value, including:

[0142] The target loss value is determined as the sum of the first loss value and the target product, the target product is the product between the second loss value and the target weight, and the target weight is associated with the ranking position of the Mth training sample in the K training samples.

[0143] In this embodiment, the electronic device determines the sum of the first loss value and the target product as the target loss value, and the target weight in the target product is associated with the sorting position of the Mth training sample in the K training samples, so that the target weight corresponding to the second loss value can be dynamically adjusted according to the training process, making the fusion processing of the first loss value and the second loss value more reasonable, further improving the accuracy of the battery SOH prediction model.

[0144] The target weight is associated with the ranking position of the Mth training sample among the K training samples. The target weight may be larger as the ranking position is later. For example, the target weight may be the ratio of M to K.

[0145] Alternatively, in some embodiments, the target weight may be M minus K / 2 divided by K / 2. In this case, the target loss value may be calculated using the following formulas (1) and (2):

[0146] total_loss = loss_1 + weight * loss_2 (1)

[0147] weight = (epoch - num_epochs / 2) / (num_epochs / 2) (2)

[0148] Among them, total_loss represents the target loss value; loss_1 represents the first loss value; loss_2 represents the second loss value; weight represents the target weight; epoch represents the ranking position of the Mth training sample among the K training samples, that is, M; num_epochs represents the total number of training samples, that is, K.

[0149] It should be noted that the above-mentioned model to be trained can be a machine learning model of any model structure.

[0150] In some embodiments, the model structure of the above-mentioned model to be trained includes:

[0151] Input layer: The input layer is a fully connected layer with 64 neurons. The input dimension of the training function is the second dimension, and its activation function is the rectified linear unit.

[0152] The first hidden layer is a fully connected layer with 16 neurons, and its activation function is the rectified linear unit;

[0153] The second hidden layer is a fully connected layer with 8 neurons, and its activation function is the rectified linear unit;

[0154] The third hidden layer is a fully connected layer with 4 neurons, and its activation function is the rectified linear unit;

[0155] The output layer is a fully connected layer with 1 neuron, and its activation function is a linear function.

[0156] In this embodiment, by setting the model structure of the above-mentioned model to be trained, the model structure is made more suitable for the characteristics of battery data, so that the model to be trained has better learning and generalization capabilities, and the accuracy of the battery SOH prediction model is further improved.

[0157] Of course, the above-mentioned model to be trained can be a machine learning model with other model structures. For example, the above-mentioned input layer can have 256 neurons, the first hidden layer can have 64 neurons, the second hidden layer can have 16 neurons, the third hidden layer can have 4 neurons, the output layer can have 1 neuron, and so on.

[0158] Step S103: When the target loss value meets the preset iteration stop condition, a battery SOH prediction model is trained.

[0159] In this step, the above-mentioned target loss value satisfies the preset iteration stopping condition. It can be that the target loss value of the current training tends to be unchanged from the target loss value of the previous training, then it is determined that the target loss value satisfies the preset iteration stopping condition; or, it can be that the target loss value is within the preset loss value range, then it is determined that the target loss value satisfies the preset iteration stopping condition, and so on.

[0160] In an embodiment of the present application, a training sample set is obtained, the training sample set includes multiple training samples, each training sample includes a historical characteristic value of the target physical aging characteristic of the battery and a corresponding label SOH; the training samples of the training sample set are sequentially input into the model to be trained for iterative training, and in the iterative training, the target loss value of the model to be trained in each training is determined according to the physical constraints associated with the target physical aging characteristic; when the target loss value meets the preset iteration stop condition, the battery SOH prediction model is trained. In this way, the loss value of the model to be trained during the training process can be affected based on the physical constraints associated with the target physical aging characteristic, so that the trained battery SOH prediction model is more suitable for the actual application scenario of the battery, thereby improving the accuracy of the battery SOH prediction model.

[0161] Figure 2 The flowchart of the embodiment of the method for predicting the SOH of a battery provided by the present application is shown in FIG. The method for predicting the SOH of a battery provided by the present application can be applied to an electronic device, which can be an onboard terminal of a vehicle or a terminal device, server, etc. independent of the vehicle. Figure 2 As shown, the method for predicting the SOH of a battery includes but is not limited to the following steps:

[0162] Step S201: obtaining a current characteristic value of a target physical aging characteristic of a battery to be tested;

[0163] Step S202: Input the current characteristic value of the target physical aging characteristic into the battery SOH prediction model to obtain the predicted SOH of the battery to be tested.

[0164] It should be noted that the above-mentioned battery to be tested can be any battery, and it can be a battery of the same type as the battery corresponding to the training sample in the training battery SOH prediction model, or it can be a battery of a different type, which is not limited here.

[0165] The battery SOH prediction model in the method for predicting the SOH of a battery provided in an embodiment of the present application is obtained by training the battery SOH prediction model method mentioned above in the present application, and the implementation principle and beneficial effects of the method for predicting the SOH of a battery are similar to those of the method for training the battery SOH prediction model, and will not be repeated here.

[0166] The method for training a battery SOH prediction model provided in an embodiment of the present application can be executed by a device for training a battery SOH prediction model. In the embodiment of the present application, the device for training a battery SOH prediction model is used as an example to illustrate the device for training a battery SOH prediction model provided in an embodiment of the present application.

[0167] Figure 3 Schematic diagram of the structure of the device for training the battery SOH prediction model provided in the embodiment of the present application. Figure 3 As shown, the apparatus 300 for training a battery SOH prediction model of the present application includes:

[0168] The sample set acquisition module 301 is used to acquire a training sample set, where the training sample set includes multiple training samples, each of which includes a historical feature value of a target physical aging feature of a battery and a corresponding label SOH;

[0169] A training module 302 is configured to sequentially input training samples of the training sample set into the model to be trained for iterative training, and determine a target loss value of the model to be trained in each training according to physical constraints associated with the target physical aging characteristics during the iterative training;

[0170] The prediction model determination module 303 is used to train and obtain a battery SOH prediction model when the target loss value meets a preset iteration stop condition.

[0171] In some embodiments, target physical aging characteristics include:

[0172] The battery's operating characteristics of continuous constant current within the preset charging time;

[0173] The accumulated characteristic quantity of the battery throughout its life cycle.

[0174] In some embodiments, the physical constraint includes a first physical constraint; the training module 302 is specifically configured to:

[0175] When the Nth training sample is input into the to-be-trained model for training, in response to the first physical constraint, obtaining the Nth first predicted SOH pre-output by the to-be-trained model, where N is a positive integer;

[0176] If the Nth first predicted SOH exceeds a preset reference SOH range, the Nth first predicted SOH is adjusted to be within the reference SOH range, where the reference SOH range is associated with the accumulated feature quantity in the Nth training sample;

[0177] According to the adjusted Nth first predicted SOH, the target loss value of the model to be trained in the Nth training is calculated.

[0178] In some embodiments, the physical constraint includes a second physical constraint; the training module 302 is specifically configured to:

[0179] When the Mth training sample is input into the to-be-trained model for training, in response to the second physical constraint, a target number of training times satisfying a preset condition in the M training times is obtained, where the preset condition is that a change trend of the second predicted SOH in two previous and subsequent training times matches a change trend of the accumulated feature quantity, and M is a positive integer;

[0180] According to the ratio of the target number of training times to M, the target loss value of the model to be trained in the Mth training is calculated.

[0181] In some implementations, the training module 302 is specifically configured to:

[0182] Calculating a first difference in the second predicted SOH between the Mth training and the M-1th training, and calculating a second difference in the accumulated feature quantities in the training samples used in the Mth training and the M-1th training;

[0183] Determine, based on the first difference and the second difference, a matching result of a change trend of the second predicted SOH and a change trend of the accumulated feature quantity between the Mth training and the M-1th training;

[0184] generating a target symbol when the matching result indicates that the change trend of the second predicted SOH matches the change trend of the accumulated feature amount;

[0185] The target number of training times is the number of target symbols generated in M ​​training times.

[0186] In some embodiments, the second predicted SOH is obtained by adjusting the first predicted SOH output by the to-be-trained model in the Mth training in response to the first physical constraint; the training module 302 is specifically configured to:

[0187] Calculate the first loss value of the model to be trained in the Mth training round based on the second predicted SOH; and use the ratio of the target number of training rounds to M as the second loss value of the model to be trained in the Mth training round.

[0188] The first loss value and the second loss value are fused to obtain the target loss value.

[0189] In some implementations, the training module 302 is specifically configured to:

[0190] If the training sample set includes K training samples, and the Mth sample is a training sample between the K / 2th training sample and the Kth training sample, the first loss value and the second loss value are fused to obtain the target loss value, where K is an integer greater than 1.

[0191] The training module 302 is further specifically configured to:

[0192] If the Mth training sample is a training sample between the 1st training sample and the K / 2th training sample, the first loss value is determined as the target loss value.

[0193] In some implementations, the training module 302 is specifically configured to:

[0194] The target loss value is determined as the sum of the first loss value and the target product, the target product is the product between the second loss value and the target weight, and the target weight is associated with the ranking position of the Mth training sample in the K training samples.

[0195] In some embodiments, the apparatus 300 further includes:

[0196] A characteristic data acquisition module is used to obtain multiple groups of historical characteristic data of the battery, each group of historical characteristic data includes historical characteristic values ​​of multiple physical aging characteristics;

[0197] a correlation determination module, configured to determine the correlation between a plurality of physical aging characteristics based on a plurality of sets of historical characteristic data;

[0198] a screening module, configured to screen at least some of the physical aging features as target physical aging features from the multiple physical aging features based on correlations between the multiple physical aging features;

[0199] The sample set generation module is used to generate a training sample set based on the historical feature values ​​of the target physical aging feature in multiple groups of historical feature data.

[0200] The device 300 for training a battery SOH prediction model provided in an embodiment of the present application can execute the technical solution shown in the above-mentioned method embodiment for training a battery SOH prediction model. Its implementation principle and beneficial effects are similar and will not be repeated here.

[0201] The method for predicting battery SOH provided in the embodiment of the present application can be executed by a device for predicting battery SOH. In the embodiment of the present application, the method for predicting battery SOH performed by the device for predicting battery SOH is used as an example to illustrate the device for predicting battery SOH provided in the embodiment of the present application.

[0202] Figure 4 Schematic diagram of the structure of the device for predicting the SOH of a battery provided in an embodiment of the present application. Figure 4 As shown, the device 400 for predicting the SOH of a battery of the present application includes:

[0203] The characteristic value acquisition module 401 is used to obtain the current characteristic value of the target physical aging characteristic of the battery to be tested;

[0204] The prediction module 402 is configured to input the current characteristic value of the target physical aging characteristic into the battery SOH prediction model to obtain the predicted SOH of the battery to be tested.

[0205] The device 400 for predicting the SOH of a battery provided in an embodiment of the present application can execute the technical solution shown in the above-mentioned method embodiment for predicting the SOH of a battery. Its implementation principle and beneficial effects are similar and will not be repeated here.

[0206] Figure 5 This is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application.

[0207] The electronic device may include a processor 501 and a memory 502 storing computer program instructions.

[0208] Specifically, the processor 501 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0209] The memory 502 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 502 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 502 may include removable or non-removable (or fixed) media. Where appropriate, the memory 502 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 502 is a non-volatile solid-state memory.

[0210] In some embodiments, the memory 502 may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method for training a battery SOH prediction model or the method for predicting the SOH of a battery according to the present application.

[0211] The processor 501 reads and executes computer program instructions stored in the memory 502 to implement the method for training a battery SOH prediction model in the above embodiment, or implements the method for predicting the battery SOH in the above embodiment.

[0212] In one example, the electronic device may further include a communication interface 503 and a bus 510. Figure 5 As shown, the processor 501, the memory 502, and the communication interface 503 are connected via a bus 510 and communicate with each other.

[0213] The communication interface 503 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0214] Bus 510 includes hardware, software or both, and the components of online data flow metering equipment are coupled to each other. For example, but not limitation, bus can include accelerated graphics port (AGP) or other graphics bus, enhanced industry standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industry standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 510 can include one or more buses. Although the present application embodiment describes and shows specific bus, the application considers any suitable bus or interconnection.

[0215] The electronic device can execute the method of training the battery SOH prediction model or the method of predicting the battery SOH in the embodiment of the present application, thereby realizing the combination of Figures 1 to 4 The method and device for training a battery SOH prediction model, or the method and device for predicting the SOH of a battery are described.

[0216] In addition, in conjunction with the method for training a battery SOH prediction model or the method for predicting a battery SOH in the above-mentioned embodiments, embodiments of the present application further provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by a processor, the method for training a battery SOH prediction model or the method for predicting a battery SOH in the above-mentioned embodiments is implemented.

[0217] In combination with the method for training a battery SOH prediction model or the method for predicting a battery SOH in the above-mentioned embodiments, an embodiment of the present application also provides a computer program product. When the instructions in the computer program product are executed by the processor of an electronic device, the electronic device implements the method for training a battery SOH prediction model or the method for predicting a battery SOH in the above-mentioned embodiments.

[0218] In combination with the electronic device in the above embodiment, an embodiment of the present application further provides a vehicle, which includes the electronic device in the above embodiment.

[0219] The vehicle can be a private car, such as a sedan, SUV, MPV, or pickup truck. It can also be a commercial vehicle, such as a van, bus, small truck, or large trailer. It can be a gasoline vehicle or a new energy vehicle. When the vehicle is a new energy vehicle, it can be a hybrid vehicle or a pure electric vehicle.

[0220] It should be understood that the present application is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, a detailed description of known methods is omitted here. In the above embodiments, several specific steps are described and illustrated as examples. However, the method process of the present application is not limited to the specific steps described and illustrated. Those skilled in the art may make various changes, modifications, and additions, or change the order of the steps after understanding the spirit of the present application.

[0221] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium that can store or transmit information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROMs, flash memories, erasable ROMs (EROMs), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0222] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps. In other words, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0223] Aspects of the present application have been described above with reference to the flow chart and / or block diagram of the method (system) and computer program product according to the embodiments of the present application. It should be understood that each box in the flow chart and / or block diagram and the combination of each box in the flow chart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine so that these instructions executed via the processor of the computer or other programmable data processing device enable the realization of the function / action specified in one or more boxes of the flow chart and / or block diagram. This processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor or a field programmable logic circuit. It is also understood that each box in the block diagram and / or flow chart and the combination of the boxes in the block diagram and / or flow chart can also be realized by the dedicated hardware that performs the specified function or action, or can be realized by the combination of dedicated hardware and computer instructions.

[0224] The above description is only a specific embodiment of the present application. Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the scope of protection of the present application.

Claims

1. A method for training a battery SOH prediction model, characterized in that: The method comprises: Acquire a training sample set, the training sample set comprising a plurality of training samples, each of the training samples comprising a historical feature value of a target physical aging feature of a battery and a corresponding label SOH; Inputting the training samples of the training sample set into the model to be trained in sequence for iterative training, and determining the target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics during the iterative training; When the target loss value meets the preset iteration stop condition, the battery SOH prediction model is trained.

2. The method according to claim 1, characterized in that The target physical aging characteristics include: The battery has a constant current condition within the preset charging time. The accumulated characteristic quantity of the battery during its entire life cycle.

3. The method according to claim 2, characterized in that The physical constraints include a first physical constraint; The step of determining the target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics in the iterative training includes: When the Nth training sample is input into the model to be trained for training, in response to the first physical constraint, obtaining the Nth first predicted SOH pre-output by the model to be trained, where N is a positive integer; If the Nth first predicted SOH exceeds a preset reference SOH range, adjusting the Nth first predicted SOH to be within the reference SOH range, the reference SOH range being associated with the accumulated feature quantity in the Nth training sample; According to the adjusted Nth first predicted SOH, the target loss value of the model to be trained in the Nth training is calculated.

4. The method according to claim 2, characterized in that: The physical constraint includes a second physical constraint; The step of determining the target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics in the iterative training includes: When the Mth training sample is input into the model to be trained for training, in response to the second physical constraint, a target number of training times satisfying a preset condition in the M training times is obtained, wherein the preset condition is that a change trend of the second predicted SOH in two previous and subsequent training times matches a change trend of the accumulated feature quantity, and M is a positive integer; According to the ratio of the target number of training times to M, the target loss value of the model to be trained in the Mth training is calculated.

5. The method according to claim 4, characterized in that The obtaining of the target number of training times satisfying a preset condition in the M training times includes: Calculating a first difference in the second predicted SOH between the Mth training and the M-1th training, and calculating a second difference in the accumulated feature quantities in the training samples used for the Mth training and the M-1th training; Determine, according to the first difference and the second difference, a matching result of a change trend of the second predicted SOH and a change trend of the accumulated feature quantity between the M-th training and the M-1-th training; generating a target symbol in a case where the matching result indicates that the change trend of the second predicted SOH matches the change trend of the accumulated feature amount; The target number of training times is the number of target symbols generated in the M training times.

6. The method according to claim 4, characterized in that The second predicted SOH is obtained by adjusting the first predicted SOH output by the model to be trained in the Mth training in response to the first physical constraint. The step of calculating the target loss value of the model to be trained in the Mth training according to the ratio of the target number of training times to M includes: According to the second predicted SOH, a first loss value of the model to be trained in the Mth training is calculated; and a ratio of the target number of training times to M is used as a second loss value of the model to be trained in the Mth training; The first loss value and the second loss value are fused to obtain the target loss value.

7. The method according to claim 6, characterized in that The fusing the first loss value and the second loss value to obtain the target loss value includes: If the training sample set includes K training samples, and the Mth sample is a training sample between the K / 2th training sample and the Kth training sample, the first loss value and the second loss value are merged to obtain the target loss value, where K is an integer greater than 1; The step of determining the target loss value of the model to be trained in each training according to the physical constraints associated with the target physical aging characteristics in the iterative training further includes: If the Mth training sample is a training sample between the 1st training sample to the K / 2th training sample, the first loss value is determined as the target loss value.

8. The method according to claim 6, characterized in that The first loss value and the second loss value are merged to obtain the target loss value, including: The sum of the first loss value and a target product is determined as the target loss value, where the target product is the product of the second loss value and a target weight, and the target weight is associated with a sorting position of the Mth training sample among the K training samples.

9. The method according to claim 1, characterized in that: Before obtaining the training sample set, the method further includes: Acquire multiple groups of historical characteristic data of the battery, each group of the historical characteristic data comprising historical characteristic values ​​of multiple physical aging characteristics; Determining, based on the multiple sets of historical feature data, a degree of correlation between the multiple physical aging features; According to the correlation between the multiple physical aging characteristics, selecting at least part of the physical aging characteristics from the multiple physical aging characteristics as the target physical aging characteristics; The training sample set is generated based on the historical feature values ​​of the target physical aging feature in the multiple groups of historical feature data.

10. A method for predicting the SOH of a battery, characterized in that: The method comprises: Obtaining current characteristic values ​​of target physical aging characteristics of the battery to be tested; Inputting the current characteristic value of the target physical aging characteristic into the battery SOH prediction model to obtain the predicted SOH of the battery to be tested, Wherein, the battery SOH prediction model is trained by the method for training a battery SOH prediction model according to any one of claims 1 to 9.

11. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the method for training a battery SOH prediction model as described in any one of claims 1 to 9, or implements the method for predicting the SOH of a battery as described in claim 10.

12. A computer-readable storage medium, characterized in that: The computer readable storage medium has computer program instructions stored thereon; When the computer program instructions are executed by a processor, the method for training a battery SOH prediction model as described in any one of claims 1 to 9 is implemented, or the method for predicting the SOH of a battery as described in claim 10 is implemented.

13. A vehicle, characterized in that: Comprising the electronic device as claimed in claim 11.

Images