Model training method and device and battery pack health state prediction method and device

Abstract

The invention provides a model training method and device and a battery pack health state prediction method and device, and relates to the technical field of battery detection, and the model training method comprises the steps: obtaining battery parameters of a plurality of batteries in a battery pack in a plurality of time periods, the battery pack is charged in a first part of time period and discharged in a second part of time period in the plurality of time periods, the charging power is greater than or equal to the standard charging power of the battery pack, and the discharging power is greater than or equal to the standard discharging power of the battery pack; obtaining the SOH (state of health) of the battery pack after each time period is ended; and training the machine learning model by taking the battery parameter of each battery in the plurality of batteries in each time period as input and taking the SOH of the battery pack after the end of each time period as output.

Description

Technical Field

[0001] The present disclosure relates to the field of battery detection technology, and in particular to a model training method and device, and a battery pack health status prediction method and device. Background Art

[0002] Currently, most batteries used in construction machinery (eg, electric excavators, electric loaders, etc.) are in the form of battery packs, which include a plurality of batteries. Summary of the invention

[0003] In the related art, the prediction result of the state of health (SOH) of the battery pack is inaccurate.

[0004] After analysis, it was found that in the related technology, when predicting the SOH of a battery pack, the battery pack as a whole is regarded as one battery.

[0005] Further analysis found that the consistency of the batteries in the battery pack is different, that is, the attenuation of each battery in the battery pack does not change linearly at the same time. In this case, when predicting the SOH of the battery pack, treating the battery pack as a whole as a single battery will lead to inaccurate prediction results of the battery pack's state of health (SOH).

[0006] In order to solve the above problems, the embodiments of the present disclosure propose the following solutions.

[0007] According to one aspect of an embodiment of the present disclosure, a model training method is provided, comprising: obtaining battery parameters of multiple batteries in a battery pack for multiple time periods, wherein the battery pack is charged in a first part of the multiple time periods and discharged in a second part of the multiple time periods, and the charging power is greater than or equal to a standard charging power of the battery pack, and the discharging power is greater than or equal to a standard discharging power of the battery pack; obtaining the state of health SOH of the battery pack after the end of each time period; and training a machine learning model by taking the battery parameters of each battery in the multiple batteries in each time period as input and the SOH of the battery pack after the end of each time period as output.

[0008] In some embodiments, obtaining the health status SOH of the battery pack after the end of each time period includes: determining an estimated SOH value of the battery pack after the end of each time period; determining an actual SOH value of the battery pack after the end of a first time period as the SOH of the battery pack after the end of the first time period; and determining the SOH of the battery pack after the end of each time period before the first time period based on the difference between the actual SOH value and the estimated SOH value of the battery pack after the end of the first time period.

[0009] According to another aspect of an embodiment of the present disclosure, a method for predicting the health status of a battery pack is provided, comprising: acquiring battery parameters of a plurality of batteries in a battery pack actually working in a plurality of time periods, wherein the battery pack is charged in a first part of the plurality of time periods and discharged in a second part of the plurality of time periods; and inputting the battery parameters of each battery in the plurality of batteries in each time period into a machine learning model to predict the health status SOH of the battery pack after the end of the plurality of time periods.

[0010] In some embodiments, during the charging process, the battery parameters include the actual charging time in each time period within the first part of the time period, and the actual charging time is less than or equal to the corresponding time period; and / or during the discharging process, the battery parameters include the actual discharge time in each time period within the second part of the time period, and the actual discharge time is less than or equal to the corresponding time period.

[0011] In some embodiments, the battery parameters further include open circuit voltage.

[0012] In some embodiments, during the charging process, the battery parameters further include charging current; and during the discharging process, the battery parameters further include discharging current.

[0013] In some embodiments, the battery parameters also include temperature.

[0014] In some embodiments, the battery parameters are obtained through a convolutional neural network (CNN), and / or the machine learning model includes a long short-term memory network (LSTM).

[0015] In some embodiments, the machine learning model applies an attention mechanism, wherein, among the multiple time periods, a later time period in the attention mechanism has a greater weight on the battery parameter.

[0016] In some embodiments, for each time period, the battery parameter of each battery is an average value of the battery parameter of each battery at multiple moments in the time period.

[0017] According to another aspect of an embodiment of the present disclosure, a model training device is provided, comprising: a module configured to execute the model training method described in any one of the above embodiments.

[0018] According to another aspect of the embodiments of the present disclosure, a device for predicting the health status of a battery pack includes: a module configured to execute the method for predicting the health status of a battery pack described in any one of the above embodiments.

[0019] According to another aspect of an embodiment of the present disclosure, there is provided an electronic device, comprising: a memory; and a processor coupled to the memory, configured to execute the method described in any one of the above embodiments based on instructions stored in the memory.

[0020] According to another aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, comprising computer program instructions, wherein when the computer program instructions are executed by a processor, the steps of the method described in any one of the above embodiments are implemented.

[0021] According to another aspect of the embodiments of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the steps of the method described in any one of the above embodiments are implemented.

[0022] In the embodiment of the present disclosure, for multiple batteries in a battery pack, battery parameters of the multiple batteries during charging and discharging in multiple time periods are obtained, and the battery parameters of each battery in the multiple batteries in each time period are used as input, and the SOH of the battery pack after each time period is used as output to train a machine learning model.

[0023] In this way, the consistency of the batteries in the battery pack can be taken into account when training the machine learning model, which helps to improve the accuracy of the trained machine learning model's prediction results for the battery pack's SOH.

[0024] The technical solution of the present disclosure is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 It is a flowchart of a method for training a model according to some embodiments of the present disclosure.

[0027] Figure 2 is a schematic diagram of an input structure of battery parameters according to some embodiments of the present disclosure.

[0028] Figure 3 It is a flowchart of a method for predicting the health status of a battery pack according to some embodiments of the present disclosure.

[0029] Figure 4 It is a schematic diagram of the structure of a training device for a model according to some embodiments of the present disclosure.

[0030] Figure 5 is a schematic diagram of the structure of a battery pack health status prediction device according to some embodiments of the present disclosure.

[0031] Figure 6 is a schematic diagram of the structure of an electronic device according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure.

[0033] Unless specifically stated otherwise, the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure.

[0034] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not necessarily drawn according to the actual proportional relationship.

[0035] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0036] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0037] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0038] In addition, in the description of the present disclosure, the terms "first", "second", "third", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance and order. Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that such operations be performed in the particular order shown or in a sequential order, or requiring that all illustrated operations be performed to achieve the desired results. In some cases, multitasking and parallel processing can be advantageous.

[0039] Figure 1 It is a flowchart of a method for training a model according to some embodiments of the present disclosure.

[0040] In step 102 , battery parameters of the multiple batteries in the battery pack are obtained in multiple time periods.

[0041] Here, the battery pack is charged in the first part of the multiple time periods and discharged in the second part of the time period, the charging power is greater than or equal to the standard charging power of the battery pack, and the discharging power is greater than or equal to the standard discharging power of the battery pack.

[0042] That is to say, in some of these multiple time periods, the battery pack is charged at a power greater than or equal to the standard charging power of the battery pack, and in other time periods, the battery pack is discharged at a power greater than or equal to the standard discharging power of the battery pack.

[0043] The battery pack is, for example, a new battery pack. By charging and discharging the new battery pack, the battery parameters of each battery in the new battery pack in multiple time periods can be obtained.

[0044] As some embodiments, the battery parameters can be stored in the cloud platform for subsequent use. For example, the battery parameters can be input from the cloud platform into the machine learning model mentioned later.

[0045] The following text will introduce how to determine the first time period and the second time period in multiple time periods, and how to determine the charging power and the discharging power in combination with some embodiments.

[0046] In step 104 , the SOH of the battery pack after each of the multiple time periods ends is obtained.

[0047] As some embodiments, the SOH of the battery pack may be measured after each time period. For example, the SOH of the battery pack may be determined using an ampere-hour integration method.

[0048] As some other implementations, the SOH of the battery pack may be estimated after each time period. For example, for each time period in the first time period, the SOH of the battery pack may be estimated based on the charging power and charging time of the battery pack in combination with historical experience; for another example, for each time period in the second time period, the SOH of the battery pack may be estimated based on the discharge power and discharge time of the battery pack in combination with historical experience.

[0049] In some embodiments, the multiple time periods are of the same length.

[0050] In this way, the time length of each time period in multiple time periods is the same as each other, and the battery parameters can be obtained periodically, which helps the machine learning model to converge better during the training process and helps to improve the efficiency of training the machine learning model.

[0051] In step 106, the battery parameters of each battery in the multiple batteries in each time period are taken as input, and the SOH of the battery pack after each time period is taken as output, and the machine learning model is trained.

[0052] That is to say, the battery parameters of each battery in each time period and the SOH of the battery pack after the end of the time period are used as training data for the machine learning model.

[0053] In the above embodiment, for multiple batteries in a battery pack, battery parameters of the multiple batteries during charging and discharging in multiple time periods are obtained, and the battery parameters of each battery in the multiple batteries in each time period are used as input, and the SOH of the battery pack after each time period is used as output to train the machine learning model.

[0054] In this way, the consistency of each battery in the battery pack can be taken into account when training the machine learning model, which helps to improve the accuracy of the trained machine learning model's prediction results for the battery pack's SOH.

[0055] In some embodiments, at least one of a first portion of the time period for charging the battery pack among the plurality of time periods and a second portion of the time period for discharging the battery pack among the plurality of time periods is determined in a random manner.

[0056] As some implementation modes, the first part of the time period and the second part of the time period may be determined by a random walk method.

[0057] In this way, by randomly determining at least one of the first time period and the second time period in multiple time periods, various conditions of battery pack charging and / or discharging can be simulated, thereby obtaining battery parameters under various conditions, which helps to further improve the accuracy of the trained machine learning model's prediction results for the battery pack's SOH.

[0058] In some embodiments, the first part of the time period and the second part of the time period are determined randomly. In this way, by randomly determining the first part of the time period and the second part of the time period in multiple time periods, various charging and discharging conditions of the battery pack can be simulated, thereby obtaining battery parameters under more conditions, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0059] In some embodiments, at least one of the charging power in each time period in the first part of the time period and the discharging power in each time period in the second part of the time period is randomly set.

[0060] As some implementation modes, random setting is performed by using a random walk method.

[0061] In this way, by randomly setting the charging power in each time period in the first time period and / or the discharging power in each time period in the second time period, the battery parameters of the battery pack under a variety of charging powers and / or a variety of discharging powers can be obtained, which helps to further improve the accuracy of the trained machine learning model's prediction results for the battery pack's SOH.

[0062] In some embodiments, the charging power of each time period in the first part of the time period and the discharging power of each time period in the second part of the time period are randomly set.

[0063] In this way, by randomly setting the charging power and discharging power of the battery pack, the battery parameters of the battery pack under a variety of charging powers and a variety of discharging powers can be obtained, which helps to further improve the accuracy of the trained machine learning model's prediction results for the battery pack's SOH.

[0064] The inventors noticed that if it is necessary to obtain the SOH of the battery pack after each time period, if the SOH of the battery pack is measured after each time period, it will take a lot of time and effort, and the estimated SOH of the battery pack is often not accurate enough.

[0065] In this regard, the present disclosure proposes the following technical solutions in combination with some embodiments.

[0066] In some embodiments, an estimated SOH value of the battery pack after each time period is determined, and an actual SOH value of the battery pack after a first time period is determined as the SOH of the battery pack after the first time period; and based on the difference between the actual SOH value and the estimated SOH value of the battery pack after the first time period, the SOH of the battery pack after each time period before the first time period is determined.

[0067] It should be understood that the actual SOH value of the battery pack after the first time period ends is the SOH of the battery pack measured after the first time period ends.

[0068] As some implementations, when the actual value of the SOH is greater than the estimated value of the SOH, the estimated values ​​of the SOH of the battery pack after each time period before the first time period ends may be increased to determine the SOH of the battery pack.

[0069] For example, the amount of increase may be the difference between the actual value of SOH and the estimated value of SOH; or, for another example, the difference between the actual value of SOH and the estimated value of SOH may be input into a trained numerical correction model to obtain the amount of increase in the estimated value of SOH. Here, the output of the trained numerical correction model is the amount of increase in the estimated value of SOH.

[0070] As some other implementations, when the actual value of the SOH is less than the estimated value of the SOH, the estimated values ​​of the SOH of the battery pack after each time period before the first time period may be reduced to determine the SOH of the battery pack.

[0071] For example, the amount of reduction may be the absolute value of the difference between the actual value of SOH and the estimated value of SOH; for another example, the difference between the actual value of SOH and the estimated value of SOH may be input into a trained numerical correction model to obtain the amount of reduction in the estimated value of SOH. Here, the output of the trained numerical correction model is the amount of reduction in the estimated value of SOH.

[0072] In the above embodiment, the estimated SOH value of the battery pack after each time period is determined, the actual SOH value of the battery pack after the first time period is used as the SOH of the battery pack after the first time period, and the SOH of the battery pack after each time period before the first time period is determined based on the difference between the actual SOH value and the estimated SOH value.

[0073] In this way, the number of times the battery pack's SOH is measured can be reduced, saving time and effort. Moreover, by correcting the battery pack's SOH based on the difference between the actual SOH value and the estimated SOH value, the accuracy of the battery pack's SOH can be improved, which helps to further improve the accuracy of the trained machine learning model's prediction results for the battery pack's SOH.

[0074] In some embodiments, during charging of the battery pack, the battery parameters serving as input to the machine learning model include the actual charging time in each time period within a first part of the time period, and / or during discharging of the battery pack, the battery parameters serving as input to the machine learning model include the actual discharge time in each time period within a second part of the time period.

[0075] Here, the actual charging time is less than or equal to the corresponding time period, and the actual discharging time is less than or equal to the corresponding time period.

[0076] That is to say, for each time period in the first time period, the battery parameters of the battery pack include the actual charging time in the time period; for each time period in the second time period, the battery parameters of the battery pack include the actual discharging time in the time period.

[0077] It should be understood that the actual charging time of the battery pack refers to the time it takes for the power of the battery pack to increase, and the actual discharging time of the battery pack refers to the time it takes for the power of the battery pack to be consumed.

[0078] For example, assuming that the battery pack is charged for 5 minutes, and the battery pack is fully charged at the 3rd minute, then the actual charging time is 3 minutes.

[0079] For another example, assuming that the battery pack is discharged for 5 minutes, and the power of the battery pack is exhausted at the 3rd minute, then the actual discharge time is 3 minutes.

[0080] In the above embodiment, when the battery pack is charged, the battery parameters include the actual charging time of the battery pack and / or when the battery pack is discharged, the battery parameters include the actual discharge time of the battery pack. In this way, the actual charging time and / or the actual discharge time of the battery pack can be considered when training the machine learning model, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0081] In some embodiments, during the charging of the battery pack during the model training process, the battery parameters obtained include one or more of the open circuit voltage (OCV), the charging current, the temperature of the battery, and the actual charging time in each time period within the first part of the time period.

[0082] In some embodiments, during the discharge of the battery pack in the model training process, the battery parameters obtained include one or more of the open circuit voltage, the discharge current, the temperature of the battery, and the actual discharge time in each time period in the second time period.

[0083] In some embodiments, during charging of the battery pack, the battery parameters include actual charging time in each time period of the first time period, and during discharging of the battery pack, the battery parameters include actual discharging time in each time period of the second time period.

[0084] In the above embodiment, when the battery pack is charged, the battery parameters include the actual charging time of the battery pack, and when the battery pack is discharged, the battery parameters include the actual discharge time of the battery pack. In this way, the actual charging time and the actual discharge time of the battery pack can be considered when training the machine learning model, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0085] In some embodiments, the battery parameters further include an open circuit voltage (OCV), where the open circuit voltage refers to the voltage between the positive and negative electrodes of the battery when the battery is not connected to a load.

[0086] It should be understood that during the process of battery charging and discharging, the battery parameters include the open circuit voltage.

[0087] In the above embodiment, the battery parameter includes the open circuit voltage of the battery. In this way, the open circuit voltage of each battery can be considered when training the machine learning model, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0088] In some embodiments, during charging of the battery pack, the battery parameters further include charging current; and during discharging of the battery pack, the battery parameters further include discharging current.

[0089] In the above embodiment, during the charging process of the battery pack, the battery parameters also include the charging current, and during the discharging process of the battery pack, the battery parameters also include the discharging current. In this way, the charging current and discharging current of each battery can be considered when training the machine learning model, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0090] In some embodiments, the battery parameters also include the temperature of the battery. It should be understood that during the process of charging and discharging the battery, the battery parameters include the temperature.

[0091] In the above embodiment, during the charging and discharging process of the battery pack, the battery parameters also include the temperature of the battery. In this way, the temperature of each battery during the charging and discharging process can be considered when training the machine learning model, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0092] Next, the preparation process before training the machine learning model is introduced in conjunction with some embodiments of the present disclosure.

[0093] In some embodiments, for multiple batteries in a battery pack, initial data of the multiple batteries in multiple time periods are acquired using sensors, and the initial data are cleaned and / or normalized. Here, the initial data includes battery parameters.

[0094] As some implementations, data cleaning includes removing outliers and supplementing missing values.

[0095] In the above embodiment, the initial data is processed by data cleaning and / or normalization. In this way, data cleaning can improve the data quality of the initial data, reduce the impact of outliers and missing values ​​on the training of the machine learning model, and help to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack; normalization can standardize the dimension of the initial data, reduce the impact of the dimension difference on the training of the machine learning model, and help to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0096] In some embodiments, the initial data is subjected to noise reduction processing.

[0097] As some implementations, a denoising autoencoder (DAE) is used to perform denoising on the initial data.

[0098] For example, the dimension of the input layer of DAE is the same as the dimension of the output layer. For example, the rectified linear function (ReLU) is used as the activation function of DAE.

[0099] In the process of training DAE, for example, mean square error (MSE) can be used as a loss function. In the process of training DAE, for example, an adaptive moment estimation (Adam) optimizer can be used as an optimizer used for training. The learning rate can be set to 0.001, for example.

[0100] In the above embodiment, by subjecting the initial data to denoising, the impact of noise on the training of the machine learning model can be reduced, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0101] In some embodiments, battery parameters are acquired through a convolutional neural network (CNN).

[0102] Next, combine Figure 2 This paper introduces how CNN obtains battery parameters.

[0103] Figure 2 is a schematic diagram of an input structure of battery parameters according to some embodiments of the present disclosure.

[0104] like Figure 2 As shown, CNN extracts the battery parameters of each battery in the battery pack from the initial data according to the time period, and reshapes the data shape of the battery parameters in each time period. Here, the time period can also be called a time step.

[0105] For example, CNN extracts parameter 1, parameter 2, parameter 3, and parameter 4 of battery 1 in time period 1 and constructs three-dimensional data in the form of (time period 1; battery 1; parameter 1, parameter 2, parameter 3, parameter 4). Here, parameter 1 to parameter 4 represent different types of battery parameters.

[0106] For example, for the process of charging a battery pack, the types of battery parameters include charging current, actual charging time, temperature and open circuit voltage; for another example, for the process of discharging a battery pack, the types of battery parameters include discharge current, actual discharge time, temperature and open circuit voltage.

[0107] It should be understood that for time period T, battery M and battery parameter N, CNN can construct three-dimensional data in the form of (time period T; battery M; battery parameter N), which represents the battery parameter N of battery M in time period T.

[0108] As some implementation methods, the activation function of CNN is the ReLU function; as some implementation methods, the convolution kernel is 3*3; as some implementation methods, the pooling method of the pooling layer is maximum pooling.

[0109] As some embodiments, a time-distributed wrapper layer is applied to wrap the convolutional layer of the CNN so that the CNN processes the convolution of each time period independently.

[0110] As some implementations, the output of the CNN may be matched to the input of the machine learning model via a flattening layer.

[0111] In the above embodiment, the battery parameters can be accurately obtained through CNN, which helps to further improve the accuracy of the prediction results of the SOH of the battery pack by the trained machine learning model.

[0112] In some embodiments, for each time period, the battery parameter of each battery is an average value of the battery parameter of each battery at multiple moments in the time period.

[0113] For example, the multiple moments are all the moments in the time period, that is, the battery parameter of each battery is the average value of each battery in the time period.

[0114] For another example, the multiple moments are partial moments of all the moments in the time period, that is, the battery parameter of each battery is the average value of the battery parameter of each battery in the partial moments in the time period. The partial moments may account for 60%, 70%, 80% or 90% of the total moments, for example.

[0115] In the above embodiment, the battery parameter of each battery is the average value of the battery parameter of each battery at multiple moments in the corresponding time period. In this way, the battery parameter of each battery in each time period can reflect the change of the battery parameter at multiple moments in the time period as a whole, which helps to further improve the accuracy of the prediction result of the trained machine learning model for the SOH of the battery pack.

[0116] In some embodiments, the machine learning model includes a long short-term memory network (LSTM). The optimizer of the LSTM is, for example, an Adam optimizer, the learning rate is, for example, 0.00001, and the loss function is, for example, MAE or RMSE.

[0117] As some implementations, the number of neurons of the LSTM is 128.

[0118] As some implementations, residual connections are applied in LSTM to reduce the possibility of overfitting of LSTM.

[0119] As some implementations, battery parameters may be acquired via CNN, and / or the machine learning model may include LSTM.

[0120] In the above embodiment, the machine learning model includes LSTM. In this way, the advantage of LSTM in processing time series data with long-term dependencies can be used to determine the dependency between the battery parameters of each battery in each time period and the SOH of the battery pack in that time period during the SOH prediction process of the battery pack, thereby helping to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0121] In some embodiments, the machine learning model applies an attention mechanism. Here, in multiple time periods, the weight of the battery parameter in the later time period in the attention mechanism is greater.

[0122] For example, the multiple time periods include time period 1 and time period 2, and time period 2 is after time period 1. In this case, the weight of the battery parameter of each battery in time period 2 is greater than the weight of the battery parameter of the battery in time period 1.

[0123] For example, the multiple time periods include time period 1, time period 2, and time period 3, time period 2 is after time period 1, and time period 3 is after time period 2. In this case, the weight of the battery parameter of each battery in time period 3 is greater than the weight of the battery parameter of the battery in time period 2, and the weight of the battery parameter of each battery in time period 2 is greater than the weight of the battery parameter of the battery in time period 1.

[0124] As some implementations, the attention mechanism is a multi-head attention mechanism, and the number of attention heads is 4. In this case, the optimizer of the attention mechanism is, for example, an Adam optimizer, the learning rate is, for example, 0.001, and the loss function is, for example, a root mean square error (RMSE).

[0125] In the above embodiment, the machine learning model applies the attention mechanism, and in multiple time periods, the weight of the battery parameters in the later time period in the attention mechanism is greater. In this way, the impact of the battery parameters in the later time period on the battery SOH can be considered more during the training process, which helps to further improve the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack after the end of multiple time periods.

[0126] In some embodiments, the machine learning model outputs the prediction result of the SOH of the battery pack through the fully connected layer and the output layer.

[0127] Here, the number of units in the fully connected layer is less than the number of neurons in the machine learning model, and the number of units in the output layer is less than the number of units in the fully connected layer. The number of units in the fully connected layer is, for example, 64, the activation function is, for example, the ReLU function, and the number of units in the output layer is, for example, 1.

[0128] In the above embodiment, the machine learning model outputs the prediction result through the output layer via the fully connected layer, and the number of units in the fully connected layer is less than the number of neurons in the machine learning model, and the number of units in the output layer is less than the number of units in the fully connected layer. In this way, the dimension of the output of the machine learning model can be gradually reduced, thereby reducing the computational burden and the potential risk of overfitting, which helps to improve the working efficiency of the trained machine learning model.

[0129] In some embodiments, the prediction result of the SOH of the battery pack of the trained machine learning model is obtained according to a preset period, and based on the prediction result and the actual value of the SOH of the corresponding battery pack, it is determined whether the machine learning model needs to be retrained. The preset period can be, for example, one week, half a month, or one month.

[0130] As some implementation methods, after the trained machine learning model is put into use, the actual SOH value of the battery pack can be measured according to a preset period, and the actual SOH value can be compared with the predicted SOH value of the battery pack predicted by the machine learning model to determine whether the machine learning model needs to be retrained.

[0131] For example, when the absolute value of the difference between the actual SOH value and the predicted SOH value is greater than a preset threshold, it is determined that the machine learning model needs to be retrained.

[0132] In the above embodiment, by regularly monitoring the prediction performance of the trained model to determine whether the machine learning model needs to be retrained, it helps to maintain the accuracy of the prediction results of the trained machine learning model on the SOH of the battery pack.

[0133] The present disclosure also proposes a method for predicting the health status of a battery pack.

[0134] Figure 3 It is a flowchart of a method for predicting the health status of a battery pack according to some embodiments of the present disclosure.

[0135] In step 302, for multiple batteries in the battery pack that is actually working, battery parameters of multiple batteries in multiple time periods are obtained. Here, the battery pack is charged in the first part of the multiple time periods and discharged in the second part of the time periods.

[0136] In some embodiments, the lengths of the multiple time periods are the same. The lengths of the multiple time periods in the health status prediction method and the multiple time periods in the model prediction method may be the same or different.

[0137] In step 304, the battery parameters of each battery in the multiple batteries in each time period are input into a machine learning model to predict the SOH of the battery pack after the multiple time periods have ended.

[0138] The machine learning model can be trained, for example, in the manner described in any of the above embodiments.

[0139] It should be understood that in the process of predicting the SOH of a battery pack, the processing of the battery parameters input into the machine learning model is similar to the processing of the battery parameters used when training the machine learning model.

[0140] For example, the SOH of the battery pack in a certain time period after the multiple time periods have ended may be predicted; for another example, the SOH of the battery pack in two or more consecutive time periods after the multiple time periods have ended may be predicted.

[0141] In the above embodiment, for multiple batteries in the battery pack, battery parameters of the multiple batteries during charging and discharging in multiple time periods are obtained, and the battery parameters of each battery in the multiple batteries in each time period are input into the machine learning model to predict the SOH of the battery pack after the multiple time periods end.

[0142] In this way, the machine learning model can take into account the consistency of each battery in the battery pack, thereby improving the accuracy of the machine learning model's prediction results for the battery pack's SOH.

[0143] In some embodiments, during the actual charging process of the battery pack, the battery parameters obtained include one or more of the open circuit voltage (OCV), the charging current, the temperature of the battery, and the actual charging time in each time period in the first part of the time period.

[0144] In some embodiments, during the discharge process of the battery pack in actual use, the battery parameters obtained include one or more of the open circuit voltage, the discharge current, the temperature of the battery, and the actual discharge time in each time period in the second time period.

[0145] In some embodiments, during charging of the battery pack, the battery parameters include actual charging time in each time period of the first time period, and during discharging of the battery pack, the battery parameters include actual discharging time in each time period of the second time period.

[0146] In the above embodiment, when the battery pack is charged, the battery parameters include the actual charging time of the battery pack, and when the battery pack is discharged, the battery parameters include the actual discharge time of the battery pack. In this way, the machine learning model can take into account the actual charging time and the actual discharge time of the battery pack, thereby further improving the accuracy of the prediction result of the machine learning model on the SOH of the battery pack.

[0147] In some embodiments, the battery parameters also include open circuit voltage (OCV).

[0148] In the above embodiment, the battery parameters include the open circuit voltage of the battery. In this way, the machine learning model can take into account the open circuit voltage of each battery, thereby further improving the accuracy of the prediction result of the machine learning model for the SOH of the battery pack.

[0149] In some embodiments, during charging of the battery pack, the battery parameters further include charging current; and during discharging of the battery pack, the battery parameters further include discharging current.

[0150] In the above embodiment, during the charging process of the battery pack, the battery parameters also include the charging current, and during the discharging process of the battery pack, the battery parameters also include the discharging current. In this way, the machine learning model can consider the charging current and discharging current of each battery, thereby further improving the accuracy of the prediction result of the machine learning model for the SOH of the battery pack.

[0151] In some embodiments, the battery parameters also include the temperature of the battery.

[0152] In the above embodiment, during the charging and discharging process of the battery pack, the battery parameters also include the temperature of the battery. In this way, the machine learning model can consider the temperature of each battery during the charging and discharging process, thereby further improving the accuracy of the prediction results of the trained machine learning model for the SOH of the battery pack.

[0153] In some embodiments, battery parameters are acquired via CNN, and / or the machine learning model includes LSTM.

[0154] In the above embodiment, the battery parameters can be accurately acquired through CNN, thereby further improving the accuracy of the prediction results of the machine learning model on the SOH of the battery pack; when the machine learning model includes LSTM, by utilizing the advantage of LSTM being good at processing time series data with long-term dependencies, the dependency between the battery parameters of each battery in each time period and the SOH of the battery pack in that time period can be determined in the SOH prediction process of the battery pack, thereby further improving the accuracy of the prediction results of the machine learning model on the SOH of the battery pack.

[0155] In some embodiments, the machine learning model applies an attention mechanism. Here, among multiple time periods, the weight of the battery parameter in the later time period in the attention mechanism is greater.

[0156] In the above embodiment, the machine learning model applies the attention mechanism, and in multiple time periods, the weight of the battery parameters in the later time period in the attention mechanism is greater. In this way, the impact of the battery parameters in the later time period on the battery SOH can be considered more during the prediction process, thereby further improving the accuracy of the prediction results of the machine learning model for the SOH of the battery pack after the end of multiple time periods.

[0157] In some embodiments, for each time period, the battery parameter of each battery is an average value of the battery parameter of each battery at multiple moments in the time period.

[0158] In the above embodiment, the battery parameter of each battery is the average value of the battery parameter of each battery at multiple moments in the corresponding time period. In this way, the battery parameter of each battery in each time period can reflect the change of the battery parameter at multiple moments in the time period as a whole, thereby further improving the accuracy of the prediction result of the machine learning model for the SOH of the battery pack.

[0159] In some embodiments, the battery parameters of each of the multiple batteries in each time period can be input into the machine learning model in real time to predict the SOH of the battery pack after the multiple time periods have ended. In this way, the real-time prediction result of the machine learning model for the SOH of the battery pack can be improved.

[0160] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0161] In some embodiments, the model training device includes a module for executing the model training method of any one of the above embodiments.

[0162] Figure 4 It is a schematic diagram of the structure of a training device for a model according to some embodiments of the present disclosure.

[0163] like Figure 4 As shown, the model training device includes a first acquisition module 401, a second acquisition module 402 and a training module 403.

[0164] The first acquisition module 401 is configured to acquire battery parameters of multiple batteries in multiple time periods for multiple batteries in the battery pack. Here, the battery pack is charged in a first part of the multiple time periods and discharged in a second part of the multiple time periods, and the charging power is greater than or equal to the standard charging power of the battery pack, and the discharging power is greater than or equal to the standard discharging power of the battery pack.

[0165] The second acquisition module 402 is configured to acquire the SOH of the battery pack after each time period ends.

[0166] The training module 403 is configured to train the machine learning model by taking the battery parameters of each battery in the multiple batteries in each time period as input and the SOH of the battery pack after each time period as output.

[0167] In some embodiments, the model training device may further include other modules to execute the model training method of any of the above embodiments.

[0168] In some embodiments, the health status prediction device of a battery pack includes a module for executing the health status prediction method of a battery pack of any one of the above embodiments.

[0169] Figure 5 is a schematic diagram of the structure of a battery pack health status prediction device according to some embodiments of the present disclosure.

[0170] like Figure 5 As shown, the battery pack health status prediction device includes an acquisition module 501 and a prediction module 502 .

[0171] The acquisition module 501 is configured to acquire battery parameters of multiple batteries in multiple time periods for multiple batteries in the actual working battery pack. Here, the battery pack is charged in the first part of the multiple time periods and discharged in the second part of the multiple time periods.

[0172] The prediction module 502 is configured to input the battery parameters of each battery in the plurality of batteries in each time period into the machine learning model to predict the SOH of the battery pack after the plurality of time periods have ended.

[0173] In some embodiments, the device for predicting the health status of a battery pack may further include other modules to execute the method for predicting the health status of a battery pack of any of the above embodiments.

[0174] Figure 6 is a schematic diagram of the structure of an electronic device according to some embodiments of the present disclosure.

[0175] In some embodiments, Figure 6 As shown, the electronic device 600 includes a memory 601 and a processor 602 coupled to the memory 601, and the processor 602 is configured to execute the method of any of the aforementioned embodiments based on the instructions stored in the memory 601. As some implementation methods, the electronic device is a model training device of any of the aforementioned embodiments or a health status prediction device of a battery pack. Accordingly, the processor 602 is configured to execute the model training method or health status prediction method of any of the aforementioned embodiments based on the instructions stored in the memory 601.

[0176] The memory 601 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application program, a boot loader, and other programs.

[0177] In some embodiments, the electronic device 600 may further include an input / output interface 603, a network interface 604, a storage interface 605, etc. The input / output interface 603, the network interface 604, the storage interface 605, and the memory 601 and the processor 602 may be connected, for example, via a bus 606. The input / output interface 603 provides a connection interface for input / output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 604 provides a connection interface for various networked devices. The storage interface 605 provides a connection interface for external storage devices such as an SD card and a USB flash drive.

[0178] The embodiment of the present disclosure further provides a computer-readable storage medium, including computer program instructions, which implement the steps of the method of any one of the above embodiments when executed by a processor.

[0179] The embodiments of the present disclosure further provide a computer program product, including a computer program, which implements the steps of the method of any one of the above embodiments when executed by a processor.

[0180] So far, various embodiments of the present disclosure have been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0181] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0182] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present disclosure. It should be understood that the functions specified in one or more processes in the flowchart and / or one or more blocks in the block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions specified in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0183] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0184] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0185] Although some specific embodiments of the present disclosure have been described in detail by way of examples, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. It should be understood by those skilled in the art that the above embodiments may be modified or some technical features may be replaced by equivalents without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A model training method, comprising: For multiple batteries in a battery pack, battery parameters of the multiple batteries in multiple time periods are obtained, wherein the battery pack is charged in a first part of the multiple time periods and discharged in a second part of the multiple time periods, the charging power is greater than or equal to the standard charging power of the battery pack, and the discharging power is greater than or equal to the standard discharging power of the battery pack; Obtaining the state of health (SOH) of the battery pack after each time period ends; and The machine learning model is trained by taking the battery parameters of each battery in the multiple batteries in each time period as input and the SOH of the battery pack after each time period as output.

2. The method according to claim 1, wherein: The obtaining of the health status SOH of the battery pack after each time period includes: Determine an estimated SOH value of the battery pack after each time period; Determining an actual SOH value of the battery pack after a first time period ends as the SOH of the battery pack after the first time period ends; and The SOH of the battery pack after each time period before the first time period ends is determined according to the difference between the actual SOH value and the estimated SOH value of the battery pack after the first time period ends.

3. A method for predicting the health status of a battery pack, comprising: For a plurality of batteries in a battery pack that is actually working, obtaining battery parameters of the plurality of batteries in a plurality of time periods, wherein the battery pack is charged in a first part of the time period and discharged in a second part of the time period; and The battery parameters of each battery in the multiple batteries in each time period are input into a machine learning model to predict the health state SOH of the battery pack after the multiple time periods end.

4. The method according to any one of claims 1 to 3, wherein: During the charging process, the battery parameters include an actual charging time in each time period within the first part of the time period, and the actual charging time is less than or equal to the corresponding time period; and / or During the discharging process, the battery parameters include an actual discharging time in each time period within the second time period, and the actual discharging time is less than or equal to the corresponding time period.

5. The method according to claim 4, wherein: The battery parameters also include open circuit voltage.

6. The method according to claim 4, wherein: During the charging process, the battery parameters also include charging current; and During the discharging process, the battery parameters also include the discharging current.

7. The method according to claim 4, wherein: The battery parameters also include temperature.

8. The method according to any one of claims 1 to 3, wherein: The battery parameters are obtained through a convolutional neural network (CNN), and / or the machine learning model includes a long short-term memory (LSTM) network.

9. The method according to any one of claims 1 to 3, wherein: The machine learning model applies an attention mechanism, wherein, among the multiple time periods, the weight of the battery parameter in the later time period in the attention mechanism is greater.

10. The method according to any one of claims 1 to 3, wherein: For each time period, the battery parameter of each battery is an average value of the battery parameter of each battery at multiple moments in the time period.

11. A model training device, comprising: A module configured to execute the method according to any one of claims 1-2 and 4-10.

12. A battery pack health status prediction device, comprising: A module configured to execute the method according to any one of claims 3 to 10.

13. An electronic device, comprising: Memory; as well as A processor coupled to the memory, configured to execute the method according to any one of claims 1 to 10 based on instructions stored in the memory.

14. A computer-readable storage medium comprising a computer program, wherein: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

15. A computer program product comprising a computer program, wherein: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

Images