Battery cell health state prediction method, detection model training method, device, equipment, storage medium and program product

Abstract

The invention relates to a cell health state prediction method, a detection model training method, devices, equipment, a storage medium and a program product, and the method comprises the steps: obtaining the charging condition data of a single charging process of electric equipment, and determining the cell type and cell aging data of each cell in the electric equipment according to the charging condition data; and further, inputting the battery cell aging data of each battery cell into a preset detection model corresponding to the battery cell type, extracting a battery cell aging feature vector of each battery cell through the preset detection model, and determining a predicted health state of each battery cell based on the battery cell aging feature vector of each battery cell. The embodiment of the invention is beneficial to improving the prediction accuracy of the health state of the battery cell.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a method for predicting the health status of a battery cell, a detection model training method, a device, equipment, a storage medium, and a program product. Background Art

[0002] Lithium-ion power batteries have been widely used as power sources and energy storage devices for electrical equipment due to their high energy density, low self-discharge rate and long service life. However, lithium-ion batteries still face the problem of battery (or battery cell) performance degradation. In order to consider the safe operation of electrical equipment, it is necessary to estimate the aging degree of the battery (or battery cell). Among them, the state of health (SOH) is a common indicator to characterize the aging degree of the battery (or battery cell).

[0003] In the related art, the health status of the battery cell is evaluated by directly measuring characteristic parameters such as battery capacity, internal resistance or impedance. However, the related art is highly sensitive to environmental factors, and the estimated results of the health status of the battery cell in the actual power-consuming equipment may not be accurate. Summary of the invention

[0004] Based on this, it is necessary to provide a battery cell health status prediction method, detection model training method, device, equipment, storage medium and program product that can improve the prediction accuracy of the battery cell health status in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for predicting the health status of a battery cell, the method comprising:

[0006] Obtain charging condition data of a single charging process of an electric device;

[0007] Determine the cell type and cell aging data of each cell in the electrical equipment according to the charging condition data;

[0008] Inputting the cell aging data of each cell into a preset detection model corresponding to the cell type;

[0009] The cell aging feature vectors of each battery cell are extracted respectively through a preset detection model, and the predicted health status of each battery cell is determined based on the cell aging feature vectors of each battery cell.

[0010] In some embodiments, the preset detection model includes a variable-length sequence feature extraction network and a battery cell health state prediction network, and the battery cell aging feature vectors of each battery cell are respectively extracted through the preset detection model, and the predicted health state of each battery cell is determined based on the battery cell aging feature vectors of each battery cell, including:

[0011] The cell aging data of each cell is respectively input into the variable length sequence feature extraction network to obtain the cell aging feature vector of each cell;

[0012] The cell aging feature vector of each cell is input into the cell health status prediction network to obtain the predicted health status of each cell.

[0013] In some embodiments, the cell aging data of each cell is input into a preset detection model corresponding to the cell type, including:

[0014] Inputting the cell aging data and auxiliary charging condition data of each cell into a preset detection model corresponding to the cell type respectively; wherein the auxiliary charging condition data of the cell is used to indicate the initial electrical parameters and / or temperature parameters of the cell in a single charging process;

[0015] Correspondingly, the cell aging data of each cell is input into the variable-length sequence feature extraction network to obtain the cell aging feature vector of each cell, including:

[0016] For each battery cell, the battery cell aging data and auxiliary charging condition data of the battery cell are input into a variable-length sequence feature extraction network to obtain a battery cell aging feature vector of the battery cell.

[0017] In some embodiments, determining the cell type and cell aging data of each cell in the electric device according to the charging condition data includes:

[0018] Determine the cell type and cell charging data of each cell in the electric device according to the charging condition data;

[0019] Data preprocessing is performed on the battery cell charging data of each battery cell to obtain the battery cell aging data of each battery cell.

[0020] In some embodiments, the method further comprises:

[0021] The auxiliary charging condition data of each battery cell is determined according to the charging condition data.

[0022] In a second aspect, the present application also provides a detection model training method, the method comprising:

[0023] Acquire multiple training samples corresponding to each battery cell type; wherein the training samples corresponding to the battery cell type include: battery cell health status training labels and corresponding battery cell aging training data;

[0024] The initial detection model is trained according to a plurality of training samples corresponding to each battery cell type to obtain a preset detection model corresponding to each battery cell type.

[0025] In some embodiments, the initial detection model is trained according to a plurality of training samples corresponding to each battery cell type to obtain a preset detection model corresponding to each battery cell type, including:

[0026] For each cell type, the cell aging training data in the training samples corresponding to the cell type is input into the initial detection model to obtain the training health status output by the initial detection model;

[0027] Adjust the parameters in the initial detection model according to the cell health status training labels and training health status in the training samples to obtain an updated initial detection model;

[0028] Other training samples corresponding to the battery cell type are used as updated training samples, and the step of inputting the battery cell aging training data in the training samples corresponding to the battery cell type into the initial detection model is returned to execute, until the preset training end conditions are met, and the updated initial detection model is used as the preset detection model corresponding to the battery cell type.

[0029] In some embodiments, the cell aging training data in the training samples corresponding to the cell type is input into the initial detection model to obtain the training health status output by the initial detection model, including:

[0030] Inputting the cell aging training data in the training samples corresponding to the cell type into the initial detection model;

[0031] A cell aging training feature vector corresponding to the cell aging training data is extracted through the initial detection model, and a training health state corresponding to the training sample is determined based on the cell aging training feature vector corresponding to the cell aging training data.

[0032] In some embodiments, the initial detection model includes an initial variable-length sequence feature extraction network and an initial cell health state prediction network, extracting a cell aging training feature vector corresponding to the cell aging training data through the initial detection model, and determining a training health state corresponding to the training sample based on the cell aging training feature vector corresponding to the cell aging training data, including:

[0033] Inputting the cell aging training data in the training samples corresponding to the cell type into the initial variable-length sequence feature extraction network to obtain the cell aging training feature vector corresponding to the cell aging training data;

[0034] The cell aging training feature vector is input into the initial cell health state prediction network to obtain the training health state corresponding to the training sample.

[0035] In some embodiments, the training samples corresponding to the cell type further include: auxiliary charging condition training data corresponding to the cell health status training label, and the cell aging training data in the training samples corresponding to the cell type is input into the initial detection model, including:

[0036] Inputting the cell aging training data and auxiliary charging condition training data in the training samples corresponding to the cell type into the initial detection model;

[0037] Correspondingly, the cell aging training data in the training samples corresponding to the cell type is input into the initial variable-length sequence feature extraction network to obtain the cell aging training feature vector corresponding to the cell aging training data, including:

[0038] The cell aging training data and the auxiliary charging condition training data are input into the initial variable-length sequence feature extraction network to obtain the cell aging training feature vector corresponding to the cell aging training data.

[0039] In some embodiments, obtaining a plurality of training samples corresponding to each battery cell type includes:

[0040] For each battery cell type, obtain multiple battery cell health status training labels corresponding to the battery cell type;

[0041] Determine, according to each battery cell health status training label, a plurality of battery cell aging training data corresponding to each battery cell health status training label;

[0042] According to the training labels of the health status of each battery cell and the corresponding multiple battery cell aging training data, multiple training samples corresponding to the battery cell type are constructed.

[0043] In some embodiments, obtaining multiple battery cell health status training labels corresponding to the battery cell types includes:

[0044] Screening and processing the historical charging condition data of multiple candidate battery cells corresponding to the battery cell type according to the preset charging voltage condition to obtain the historical charging condition data of multiple first reference battery cells that meet the preset charging voltage condition;

[0045] For each first reference battery cell, the actual capacity of the first reference battery cell is determined based on the historical charging condition data of the first reference battery cell, and the battery health status training label of the first reference battery cell is determined based on the ratio of the actual capacity of the first reference battery cell to the rated capacity corresponding to the battery cell type.

[0046] In some embodiments, determining the actual capacity of the first reference battery cell according to the historical charging condition data of the first reference battery cell includes:

[0047] Determine the charging capacity of the first reference battery cell in a first interval of a preset voltage range according to the historical charging condition data of the first reference battery cell;

[0048] The actual capacity of the first reference battery cell is determined according to the product of the charging capacity in the first interval and the charging capacity correction coefficient corresponding to the battery cell type.

[0049] In some embodiments, the method further comprises:

[0050] Screening and processing the historical charging condition data of multiple candidate cells corresponding to the cell type according to the preset reference cell condition to obtain the historical charging condition data of multiple second reference cells that meet the preset reference cell condition;

[0051] Determine an average charging capacity of the plurality of second reference cells in a preset voltage range according to the historical charging condition data of the plurality of second reference cells;

[0052] The charging power correction coefficient corresponding to the battery cell type is determined based on the ratio of the average charging power in the interval to the rated capacity corresponding to the battery cell type.

[0053] In some embodiments, determining the average charging power of the plurality of second reference cells in a preset voltage range according to the historical charging condition data of the plurality of second reference cells includes:

[0054] For each second reference battery cell, determining a second interval charging capacity of the second reference battery cell in a preset voltage range according to historical charging condition data of the second reference battery cell;

[0055] The average value of the second interval charging power of each second reference battery cell is calculated to obtain the interval charging power average value.

[0056] In some embodiments, a plurality of cell aging training data corresponding to each cell health status training label is determined according to each cell health status training label, including:

[0057] For each battery cell health status training label, determine a target battery cell reference parameter corresponding to the first reference battery cell according to the historical charging condition data of the first reference battery cell corresponding to the battery cell health status training label;

[0058] Obtain other historical charging condition data in which the difference between the cell reference parameter of the first reference cell and the target cell reference parameter meets the preset cell reference condition, and determine multiple cell aging training data corresponding to the cell health status training label based on the other historical charging condition data.

[0059] In a third aspect, the present application also provides a device for predicting the health status of a battery cell, the device comprising:

[0060] An acquisition module is used to acquire charging condition data of a single charging process of an electric device;

[0061] A determination module, used to determine the cell type and cell aging data of each cell in the electrical device according to the charging condition data;

[0062] An input module, used to input the cell aging data of each cell into a preset detection model corresponding to the cell type;

[0063] The prediction module is used to extract the cell aging feature vector of each cell through a preset detection model, and determine the predicted health state of each cell based on the cell aging feature vector of each cell.

[0064] In a fourth aspect, the present application also provides a detection model training device, the device comprising:

[0065] An acquisition module is used to acquire multiple training samples corresponding to each battery cell type; wherein the training samples corresponding to the battery cell type include: battery cell health status training labels and corresponding battery cell aging training data;

[0066] The training module is used to train the initial detection model according to multiple training samples corresponding to each battery cell type to obtain a preset detection model corresponding to each battery cell type.

[0067] In a fifth aspect, the present application also provides a battery cell health status prediction device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods in the first aspect when executing the computer program.

[0068] In a sixth aspect, the present application also provides a detection model training device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods in the second aspect when executing the computer program.

[0069] In a seventh aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method of any one of the first aspect or the second aspect when the computer program is executed by a processor.

[0070] In an eighth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method of any one of the first aspect or the second aspect above.

[0071] The above-mentioned battery cell health status prediction method, detection model training method, device, equipment, storage medium and program product obtain the charging condition data of a single charging process of the electrical equipment, and determine the cell type and cell aging data of each battery cell in the electrical equipment based on the charging condition data; further, the cell aging data of each battery cell is respectively input into a preset detection model corresponding to the battery cell type, and the cell aging feature vector of each battery cell is respectively extracted through the preset detection model, and the predicted health status of each battery cell is determined based on the cell aging feature vector of each battery cell. It can be seen that in the embodiments of the present application, the cell aging data of each battery cell is respectively input into the preset detection model corresponding to the battery cell type, the cell aging feature vector of each battery cell is respectively extracted through the preset detection model, and the predicted health state of each battery cell is determined based on the cell aging feature vector of each battery cell. Since the cell aging feature vector (which belongs to a high-order feature) extracted based on the cell aging data can be used to characterize the key aging characteristics of the battery cell to determine the predicted health state of the battery cell, the information of the cell aging feature vector is richer and clearer, which is beneficial to improving the prediction accuracy of the health state of the battery cell. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Moreover, the same reference numerals are used throughout the drawings to represent the same components. In the drawings:

[0073] Figure 1 A schematic diagram of a flow chart of a method for predicting the health status of a battery cell provided in some embodiments of the present application;

[0074] Figure 2 A schematic diagram of the structure of a preset detection model provided in some embodiments of the present application;

[0075] Figure 3 A schematic diagram of a flow chart of a method for predicting the health status of a battery cell provided in some other embodiments of the present application;

[0076] Figure 4 A flowchart of a detection model training method provided in some embodiments of the present application;

[0077] Figure 5 A schematic diagram of a flow chart of a detection model training method provided in some other embodiments of the present application;

[0078] Figure 6 A schematic diagram of a flow chart of a detection model training method provided in some other embodiments of the present application;

[0079] Figure 7A flowchart of a training sample acquisition method provided in some embodiments of the present application;

[0080] Figure 8 A schematic diagram of a flow chart of a detection model training method provided in some other embodiments of the present application;

[0081] Fig. 9 A schematic diagram of a curve showing a change in the cell voltage as the charging power increases provided in an embodiment of the present application;

[0082] Fig.10 A schematic flow chart of a method for determining the actual capacity of a first reference battery cell provided in an embodiment of the present application;

[0083] Fig.11 A flow chart of a method for determining a charging power correction coefficient provided in an embodiment of the present application;

[0084] Fig.12 A schematic diagram of the overall process of the detection model training method and the battery cell health status prediction method provided in the embodiments of the present application;

[0085] Fig.13 This is a schematic diagram of the structure of a battery cell health status prediction device in some embodiments of the present application;

[0086] Fig.14 This is a schematic diagram of the structure of a detection model training device in some embodiments of the present application;

[0087] Fig.15 This is a schematic diagram of the structure of a computer device in some embodiments of the present application. DETAILED DESCRIPTION

[0088] The following embodiments of the technical solution of the present application are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0089] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the term "including" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0090] The battery cell health status prediction method, detection model training method, device, equipment, storage medium and program product in the embodiments of the present application can be applied to the application scenario of predicting the battery cell health status of electrical equipment under different charging conditions; of course, it can also be applied to other application scenarios.

[0091] Exemplarily, different charging conditions may include but are not limited to: a first charging mode, a second charging mode, and a power-consumer-power-consumer charging mode; wherein the charging speed of the first charging mode is greater than the charging speed of the second charging mode; and the power-consumer-power-consumer charging mode may refer to a mode in which one power-consumer charges another power-consumer. For example, the first charging mode may be a fast charging mode, and the second charging mode may be a slow charging mode.

[0092] It should be noted that, for the sake of convenience, the following embodiments are described by taking the electric device of the embodiment of the present application as a vehicle as an example. It should be understood that when the electric device of the embodiment of the present application is other equipment, its implementation principle and technical effect are similar.

[0093] In recent years, with the rapid growth of the practicality and diversity of electric vehicles, the market share of this new type of transportation has entered a period of rapid development and prosperity. Lithium-ion power batteries have been widely used as power sources and energy storage devices for electric vehicles and other electrical equipment due to their high energy density, low self-discharge rate and long service life. However, lithium-ion batteries still face the problem of battery (or battery cell) performance degradation, which is characterized by capacity decay or power decay. In order to consider the safe operation of electrical equipment such as electric vehicles and the cascade utilization of batteries, it is necessary to estimate the degree of aging of batteries (or batteries). Among them, the state of health (SOH) is a commonly used indicator to characterize the degree of aging of batteries (or batteries). For example, it is generally believed that when the SOH of a battery (or battery cell) is lower than 80%, the battery (or battery cell) will stop being used.

[0094] In the related technology, the health status of the battery cell is evaluated by directly measuring characteristic parameters such as battery capacity, internal resistance or impedance. However, the related technology is highly sensitive to environmental factors and is more suitable for evaluating the health status of battery cells in laboratories, but the estimated results of the health status of battery cells in actual electrical equipment may not be accurate.

[0095] In addition, in the related technology, the battery health status can also be calibrated by indirectly measuring certain process parameters that can reflect the battery capacity or internal resistance degradation, but it usually requires specific stable current or specific temperature and other conditions, and has high requirements on external factors.

[0096] In order to solve the problem in the related art that the estimation results of the health status of battery cells in actual electrical equipment may not be accurate, the present application proposes a method of respectively inputting the cell aging data of each battery cell into a preset detection model corresponding to the battery cell type, respectively extracting the cell aging feature vector of each battery cell through the preset detection model, and determining the predicted health status of each battery cell based on the cell aging feature vector of each battery cell. Since the cell aging feature vector (which is a high-order feature) extracted based on the cell aging data can be used to characterize the key aging characteristics of the battery cell to determine the predicted health status of the battery cell, the information of the cell aging feature vector is richer and clearer, which is beneficial to improving the prediction accuracy of the health status of the battery cell.

[0097] For ease of understanding, some of the names are first introduced separately in the embodiments of the present application.

[0098] The state of health SOH can be used to reflect the current aging degree of the battery (or battery cell), which can usually be the percentage of the current capacity of the battery to the factory capacity.

[0099] In the field of new energy vehicles, battery cells (or simply cells) usually refer to the individual cells that make up the battery pack, and battery usually refers to the battery pack.

[0100] The differential voltage (DV) curve is an important analytical tool in battery research, mainly used to reveal the subtle characteristics of the voltage change with the charge during the battery charging and discharging process. The horizontal axis of the DV curve can be the cell voltage, and the vertical axis can be dQ / dV; where dQ represents the change in charge, and dV represents the change in cell voltage; it can be seen that the DV curve is mainly generated by two data: charge and cell voltage.

[0101] The incremental capacity (IC) curve is also called the capacity increment curve. It is a curve drawn by measuring the corresponding relationship between small capacity changes and voltage during the charge and discharge process of electrochemical batteries such as lithium-ion batteries. The IC curve usually refers to the dQ / dV-V curve, that is, the curve of the incremental capacity as the voltage changes. Among them, the horizontal axis of the IC curve can be the battery cell voltage, and the vertical axis can be dQ / dV.

[0102] It should be understood that under different charging conditions (for example, different initial charging voltages, initial charging currents and / or initial charging temperatures), the polarization of the battery cell (for example, ohmic polarization, electrochemical polarization, and concentration polarization, etc.) is completely different, which will result in completely different voltage changes of the battery cells under different charging conditions when the same amount of electricity is charged. In other words, the curve between the charging amount and the voltage data of the same battery cell under different charging conditions (such as the curve shown in the following figure) is different. Fig. 9This will directly lead to significant changes in the DV curve or IC curve.

[0103] It should be noted that the relevant features of the DV curve or IC curve involved in the embodiments of the present application can clearly characterize the aging mechanism of the battery cell, and the curve features are accurate and effective.

[0104] In some embodiments, Figure 1 The flowchart of the method for predicting the health status of a battery cell provided in some embodiments of the present application is shown in FIG. 1 , and the method is applied to a device for predicting the health status of a battery cell as an example. Figure 1 As shown, the method of the embodiment of the present application may include the following steps:

[0105] Step S101: Acquire charging condition data of a single charging process of an electric device.

[0106] In this step, the battery cell health status prediction device can obtain the charging condition data of a single charging process of any power-consuming device to be predicted. It should be noted that any charging condition data or any historical charging condition data involved in the embodiments of the present application are charging condition data of actual market power-consuming devices.

[0107] The charging condition data in the embodiment of the present application can be used to indicate the charging data information of each battery cell in the electric device during a single charging process. Exemplarily, the charging condition data may include but is not limited to at least one of the following: an identification code of the electric device, a battery cell identification, a charging capacity of the battery cell, and a voltage of the battery cell. Among them, the battery cell identification can be used to uniquely identify the battery cell. For example, the battery cell identification may include but is not limited to the battery cell number.

[0108] It should be understood that the charging condition data may also include other charging data information, such as the initial charging current of the battery cell, the initial charging temperature of the battery cell, the charging date, the power-on operation time of the electrical equipment, and / or the factory delivery time of the electrical equipment.

[0109] It should be noted that, when the electrical device is a vehicle, the power-on operation time of the electrical device may include but is not limited to information such as vehicle mileage.

[0110] In a possible implementation, the battery cell health status prediction device may obtain charging condition data of a single charging process of any electrical device at intervals of a first preset time period.

[0111] Exemplarily, the battery cell health status prediction device can obtain the charging condition data of a single charging process of any electrical device from a database, or can obtain the charging condition data of a single charging process of any electrical device from other devices (such as a power battery big data platform, etc.).

[0112] In another possible implementation, the battery cell health status prediction device can obtain the charging condition data of a single charging process of any electrical device upon receiving a prediction instruction.

[0113] Of course, the battery cell health status prediction device can also obtain the charging condition data of a single charging process of any electrical device through other means.

[0114] Step S102: determining the cell type and cell aging data of each cell in the electrical equipment according to the charging condition data.

[0115] In this step, the battery health status prediction device can determine the battery type and battery aging data of each battery in the electric device according to the charging condition data of the single charging process of the electric device obtained in the above step S101. The battery aging data can be battery data used to characterize the battery aging mechanism.

[0116] For example, the cell type in the embodiment of the present application may vary according to the cell type, cell capacity and / or manufacturer (or supplier), etc. The cell type may include but is not limited to ternary cells or iron-lithium cells.

[0117] Exemplarily, the cell aging data may include but is not limited to: the voltage of the cell, the charge capacity change data of the cell at every preset sampling interval, and the voltage change data of the cell at every preset sampling interval. It should be understood that the cell aging data in the embodiment of the present application can be referred to as DV-like data or IC-like data (because the DV curve or IC curve contains horizontal and vertical axes and cannot be input as a whole as a sequence data).

[0118] In one possible implementation, the battery cell health status prediction device can determine the battery cell type and battery cell charging data of each battery cell in the electrical equipment based on the charging condition data, and pre-process the battery cell charging data of each battery cell to obtain the battery cell aging data of each battery cell.

[0119] Exemplarily, the battery cell health status prediction device can determine the battery cell type of each battery cell according to the battery cell identification of each battery cell in the charging condition data, and obtain the battery cell charging data of each battery cell according to the charging power and battery cell voltage of each battery cell in the charging condition data. The battery cell charging data may include but is not limited to the charging power and battery cell voltage.

[0120] Furthermore, the battery cell health status prediction device can perform data preprocessing on the battery cell charging data of each battery cell respectively to obtain the battery cell aging data of each battery cell; wherein the data preprocessing may include but is not limited to at least one of the following: data conversion processing, data balancing processing, and data downsampling processing. Among them, the data conversion processing can be used to calculate the charging power change data of the battery cell at every preset sampling interval and the voltage change data of the battery cell at every preset sampling interval, that is, to convert the battery cell charging data into the battery cell aging data. Data balancing processing can be used to avoid the influence of missing or abnormal values. Data downsampling processing can be used to reduce the computational burden.

[0121] It should be understood that for the application scenario of predicting the health status of the battery cell in the second charging mode, since the charging current is small and the charging time is long, the change between two adjacent frames of data is very small. Through data downsampling processing, downsampling can be performed without affecting the prediction of the preset detection model, so as to reduce the calculation burden. For the application scenario of predicting the health status of the battery cell in the first charging mode, data downsampling processing may not be required.

[0122] In another possible implementation, the battery cell health status prediction device may input the charging condition data into a preset extraction model to obtain the cell type and cell aging data of each battery cell in the electrical equipment output by the preset extraction model; wherein the preset extraction model may include but is not limited to a first machine learning model or a first neural network model.

[0123] Of course, the battery cell health status prediction device can also determine the battery cell type and battery cell aging data of each battery cell in the electrical equipment in other ways based on the charging condition data.

[0124] In some embodiments, the battery cell health status prediction device can also determine the auxiliary charging condition data of each battery cell according to the charging condition data, so as to further predict the battery cell health status in combination with the auxiliary charging condition data of each battery cell.

[0125] Exemplarily, the auxiliary charging condition data of the battery cell can be used to indicate the initial electrical parameters and / or temperature parameters of the battery cell during a single charging process, that is, the auxiliary charging condition data of the battery cell can be used to indicate different charging conditions. Exemplarily, the auxiliary charging condition data of the battery cell can include but is not limited to at least one of the following: the initial charging current of the battery cell, the initial charging voltage of the battery cell, and the initial charging temperature of the battery cell.

[0126] Of course, the auxiliary charging condition data of the battery cell can also use other auxiliary condition data.

[0127] Step S103: inputting the cell aging data of each cell into a preset detection model corresponding to the cell type.

[0128] In the embodiment of the present application, the battery cell health status prediction device may be preset with preset detection models corresponding to different battery cell types; wherein different battery cell types correspond to different preset detection models. For example, battery cell type 1 corresponds to preset detection model 1, and battery cell type 2 corresponds to preset detection model 2.

[0129] The preset detection model corresponding to any cell type in the embodiment of the present application can be used to indicate the corresponding relationship between the cell aging data of the cell type and the predicted health state of the corresponding cell. The preset detection model may include but is not limited to a second machine learning model or a second neural network model.

[0130] In this step, the battery cell health status prediction device can determine the corresponding preset detection model according to the battery cell type of each battery cell. Further, for each battery cell, the battery cell health status prediction device can input the battery cell aging data of the battery cell into the preset detection model corresponding to the battery cell type of the battery cell, so that the preset detection model can predict the battery cell health status.

[0131] In some embodiments, considering that the polarization of battery cells (such as ohmic polarization, electrochemical polarization, and concentration polarization, etc.) is completely different under different charging conditions, which may cause different battery cell aging conditions, the battery cell health status prediction device can input the battery cell aging data and auxiliary charging condition data of each battery cell into a preset detection model corresponding to the battery cell type.

[0132] In the embodiment of the present application, for each battery cell, the battery cell health status prediction device can input the battery cell aging data and auxiliary charging condition data of the battery cell into a preset detection model corresponding to the battery cell type of the battery cell, so that the preset detection model can predict the battery cell health status. It can be seen that by further introducing the auxiliary charging condition data to predict the battery cell health status based on the battery cell aging data, it is beneficial to further improve the accuracy of the battery cell health status prediction.

[0133] Step S104: extracting the cell aging feature vectors of each cell respectively through a preset detection model, and determining the predicted health status of each cell respectively based on the cell aging feature vectors of each cell.

[0134] In the embodiment of the present application, the cell aging feature vector of any cell can be used to characterize the key aging features of the cell, so that the health state can be accurately predicted based on the key aging features. It should be noted that the key aging features in the embodiment of the present application can be referred to as DV-like curve features or IC-like curve features.

[0135] In this step, the battery cell health status prediction device can extract the battery cell aging feature vector of each battery cell respectively through the preset detection model corresponding to the battery cell type, and determine the predicted health status of each battery cell based on the battery cell aging feature vector of each battery cell.

[0136] Exemplarily, for each battery cell, the battery cell health status prediction device can extract the battery cell aging feature vector of the battery cell through a preset detection model corresponding to the battery cell type of the battery cell, and determine the predicted health status of the battery cell based on the battery cell aging feature vector of the battery cell.

[0137] In an embodiment of the present application, the cell aging data of each battery cell is respectively input into a preset detection model corresponding to the battery cell type, and the cell aging feature vector of each battery cell is respectively extracted through the preset detection model. Since the cell aging data of each battery cell includes the cell data of each battery cell during a single charging process that can be used to characterize the cell aging mechanism, the extracted cell aging feature vector (belonging to a high-order feature) that can be used to characterize the key aging characteristics of the battery cell has richer and clearer information.

[0138] Exemplarily, after determining the predicted health status of each battery cell in the electrical equipment, the battery cell health status prediction device can send the health status detection result to the user equipment or other user terminals; wherein the health status detection result includes the predicted health status of each battery cell; other user terminals may include but are not limited to smart terminals associated with the user equipment, such as smart phones, computers or tablets.

[0139] As another example, when the battery cell health status prediction device determines the predicted health status of each battery cell in the electrical equipment and the predicted health status of any battery cell is less than a preset health status threshold, the battery cell health status prediction device can send a health status detection result to the user device or other user terminal; wherein the health status detection result includes the predicted health status of each battery cell.

[0140] In summary, in the embodiment of the present application, the charging condition data of a single charging process of an electrical device is obtained, and the cell type and cell aging data of each cell in the electrical device are determined according to the charging condition data; further, the cell aging data of each cell is respectively input into a preset detection model corresponding to the cell type, and the cell aging feature vector of each cell is respectively extracted through the preset detection model, and the predicted health state of each cell is determined based on the cell aging feature vector of each cell. It can be seen that in the embodiment of the present application, the cell aging data of each cell is respectively input into a preset detection model corresponding to the cell type, and the cell aging feature vector of each cell is respectively extracted through the preset detection model, and the predicted health state of each cell is determined based on the cell aging feature vector of each cell. Since the cell aging feature vector (belonging to a high-order feature) extracted based on the cell aging data can be used to characterize the key aging characteristics of the cell to determine the predicted health state of the cell, the information of the cell aging feature vector is richer and clearer, which is conducive to improving the prediction accuracy of the health state of the cell.

[0141] In some embodiments, Figure 2 A schematic diagram of the structure of a preset detection model provided in some embodiments of the present application, such as Figure 2 As shown, the preset detection model in the embodiment of the present application may include but is not limited to a variable-length sequence feature extraction network and a battery cell health status prediction network; wherein the variable-length sequence feature extraction network is a type of network specially designed to process input sequences of variable length. Figure 3 This is a flowchart of a method for predicting the health status of a battery cell provided in some other embodiments of the present application. The present application embodiment provides an exemplary introduction to the relevant contents of "extracting the battery cell aging feature vector of each battery cell through a preset detection model, and determining the predicted health status of each battery cell based on the battery cell aging feature vector of each battery cell" in the above step S104. Figure 3 As shown, the method of the embodiment of the present application may include the following steps:

[0142] Step S1041: inputting the cell aging data of each cell into a variable-length sequence feature extraction network to obtain a cell aging feature vector of each cell.

[0143] The variable-length sequence feature extraction network in the embodiment of the present application can be used to perform information fusion (integration) on the cell aging data of each cell to extract key aging features (for example, DV curve features or IC curve features) and output a cell aging feature vector with a fixed length, so that the features reflected by the output are closer to the DV curve or IC curve. Among them, the reason why the variable-length sequence feature extraction network outputs a cell aging feature vector with a fixed length is that the cell health status prediction network cannot adjust the network structure according to the sequence length, and usually requires a fixed input dimension parameter, so the input is data with the same dimension.

[0144] In this step, the battery cell health status prediction device can input the battery cell aging data of each battery cell into the variable-length sequence feature extraction network in the preset detection model corresponding to the battery cell type, and obtain the battery cell aging feature vector of each battery cell.

[0145] Exemplarily, for each battery cell, the battery cell health status prediction device may input the battery cell aging data of the battery cell into a variable-length sequence feature extraction network in a preset detection model corresponding to the battery cell type to obtain a battery cell aging feature vector of the battery cell.

[0146] It should be understood that in order to obtain the features that best represent the aging of the battery cells and improve the coverage of scenarios at the same time, the embodiments of the present application start from the essence of the DV curve or the IC curve, and use the battery cell aging data (DV-like data or IC-like data) corresponding to each battery cell throughout a single charge as the input of the variable-length sequence feature extraction network, and based on the powerful fitting ability of the neural network, automatically extract the battery cell aging features similar to those represented by the DV curve or the IC curve (which can accurately reflect the key features of battery cell aging) without the need to perform manual feature extraction operations.

[0147] In some embodiments, the variable-length sequence feature extraction network in the embodiments of the present application may include but is not limited to: a feature extraction unit and a self-attention unit. Among them, the feature extraction unit can be used to fuse the cell aging data of each cell to extract key aging features; the self-attention unit can be used to give higher weights to the feature information of local key areas, so that the variable-length sequence feature extraction network can better capture local key features, and can also compress long sequence inputs into fixed-length feature vectors.

[0148] Exemplarily, the feature extraction unit may include, but is not limited to: a Bi-GRU neural network, or a long short-term memory (LSTM) network model.

[0149] It should be noted that, since DV-like data or IC-like data is used as the preset detection model input in the embodiment of the present application, and the length of the battery cell aging data sequence input each time is different, a variable-length sequence feature extraction network is used (although the battery cell aging is mainly reflected in local areas such as peaks / valleys rather than the entire curve in the DV curve or IC curve, local fragment data cannot be used, because the DV curve or IC curve under different charging conditions is different. At this time, the local fragment data will change and cannot be detected. The use of complete data can accurately grasp the current curve shape).

[0150] Of course, the variable-length sequence feature extraction network in the embodiment of the present application may also include other network structures.

[0151] In some embodiments, when the cell aging data and auxiliary charging condition data of each cell are respectively input into the preset detection model corresponding to the cell type in the above step S103, for each cell, the cell health status prediction device can input the cell aging data and auxiliary charging condition data of the cell into the variable-length sequence feature extraction network to obtain the cell aging feature vector of the cell. Since the embodiment of the present application comprehensively considers the cell aging data and auxiliary charging condition data of the cell (which can be used to indicate different charging conditions), it is beneficial for the variable-length sequence feature extraction network to obtain a more accurate cell aging feature vector, thereby facilitating the cell health status prediction network to more accurately predict the predicted health status of the cell.

[0152] Step S1042: inputting the cell aging feature vector of each cell into the cell health state prediction network respectively to obtain the predicted health state of each cell.

[0153] The preset detection model corresponding to any battery cell type in the embodiments of the present application can be used to indicate the corresponding relationship between the battery cell aging feature vector of the battery cell type and the corresponding predicted health status.

[0154] In this step, the battery cell health status prediction device can input the battery cell aging feature vector of each battery cell into the battery cell health status prediction network in the preset detection model corresponding to the battery cell type to obtain the predicted health status of each battery cell.

[0155] Exemplarily, for each battery cell, the battery cell health status prediction device may input the battery cell aging feature vector of the battery cell into the battery cell health status prediction network in a preset detection model corresponding to the battery cell type of the battery cell to obtain the predicted health status of the battery cell.

[0156] Exemplarily, the preset detection model in the embodiment of the present application may include but is not limited to any of the following: Convolutional Neural Networks (CNN) model, GRU network model, LSTM network model. Among them, CNN is a type of feedforward neural network that includes convolution calculation and has a deep structure. It is one of the representative algorithms of deep learning and is more suitable for analyzing and processing high-dimensional data information.

[0157] Of course, the preset detection model in the embodiment of the present application may also include other network structures, such as a Linear network, etc. Among them, the Linear network can be used to receive input and apply a linear transformation.

[0158] Since the cell aging feature vector of any cell in the embodiment of the present application is rich and clear in information, the preset detection model in the embodiment of the present application can accurately predict the health status of the cell without having an overly complex structure.

[0159] In summary, in the embodiment of the present application, the cell aging feature vector of each cell is obtained by inputting the cell aging data of each cell into the variable length sequence feature extraction network; further, the cell aging feature vector of each cell is respectively input into the cell health state prediction network to obtain the predicted health state of each cell. It can be seen that in the embodiment of the present application, the cell aging feature vectors of the same length can be extracted from the cell aging data of different lengths under different charging conditions through the variable length sequence feature extraction network, so that the health state of the cell can be predicted through the cell health state prediction network. Since the cell aging feature vector contains rich and clear key aging features that can be used to characterize the cell, the method of predicting the health state of the cell by the cell health state prediction network based on the cell aging feature vector of the cell extracted by the variable length sequence feature extraction network (rather than the indicator that may have errors generated by other battery algorithms) is not only conducive to improving the prediction accuracy of the health state of the cell, but also the structure of the preset detection model is relatively simple, which is also conducive to improving the prediction efficiency of the health state of the cell.

[0160] In some embodiments, the embodiments of the present application illustrate relevant contents of the training process of the preset detection model corresponding to each battery cell type involved in the above embodiments. Figure 4 The following is a flow chart of a detection model training method provided in some embodiments of the present application. In the embodiments of the present application, the method is applied to a detection model training device as an example for explanation. It should be noted that the detection model training device in the embodiments of the present application and the detection model training device in the above embodiments can be the same device or different devices. Figure 4 As shown, the method of the embodiment of the present application may include the following steps:

[0161] Step S401, obtaining a plurality of training samples corresponding to each battery cell type; wherein the training samples corresponding to the battery cell type include: battery cell health status training labels and corresponding battery cell aging training data.

[0162] The cell aging training data in the embodiment of the present application may include but is not limited to the training voltage of the cell, the charging capacity change training data of the cell at each preset sampling interval, and the voltage change training data of the cell at each preset sampling interval.

[0163] The cell health status training label corresponding to the cell aging training data in the embodiment of the present application can be used to indicate the actual health status of the cell corresponding to the cell aging training data.

[0164] In this step, the detection model training device can obtain multiple training samples corresponding to each battery cell type; wherein, the training samples corresponding to any battery cell type may include but are not limited to: battery cell health status training labels and corresponding battery cell aging training data.

[0165] In a possible implementation, the detection model training device may obtain multiple training samples corresponding to each battery cell type at intervals of a second preset time duration.

[0166] In another possible implementation, the detection model training device may obtain multiple training samples corresponding to each battery cell type upon receiving a training instruction.

[0167] Of course, the detection model training device can also obtain multiple training samples corresponding to each battery cell type through other methods.

[0168] Step S402: training the initial detection model according to a plurality of training samples corresponding to each battery cell type, to obtain a preset detection model corresponding to each battery cell type.

[0169] In this step, the detection model training device can train the initial detection model according to multiple training samples corresponding to each battery cell type to obtain a preset detection model corresponding to each battery cell type. The preset detection model corresponding to any battery cell type can be used to extract the battery cell aging feature vector of each battery cell of the battery cell type, and determine the predicted health status of each battery cell based on the battery cell aging feature vector of each battery cell.

[0170] Exemplarily, for each cell type, the detection model training device may train the initial detection model according to multiple training samples corresponding to the cell type until the preset training end condition is met, and the preset detection model corresponding to the cell type may be obtained. The preset training end condition may include but is not limited to any of the following: the mean squared error loss (MSE) of the detection model is less than a preset threshold, or the number of training iterations exceeds a preset threshold. For example, the preset threshold may include but is not limited to 0.001.

[0171] It should be noted that due to the large differences between different battery cell types, a corresponding preset detection model needs to be trained for each battery cell type.

[0172] In summary, in the embodiment of the present application, multiple training samples corresponding to each battery cell type are obtained; wherein, the training samples corresponding to the battery cell type include: battery cell health status training labels and corresponding battery cell aging training data. Further, the initial detection model is trained according to the multiple training samples corresponding to each battery cell type, and the preset detection model corresponding to each battery cell type is obtained. It can be seen that since the training samples corresponding to any battery cell type in the embodiment of the present application include battery cell aging training data that can be used to characterize the battery cell aging mechanism, the method of training the prediction model based on the battery cell aging training data and the corresponding battery cell health status training labels is conducive to improving the accuracy of the preset detection model. In addition, in the embodiment of the present application, the preset detection model corresponding to different battery cell types is obtained by training the training samples corresponding to different battery cell types, so that the corresponding preset detection model can be used to predict the health status of the battery cell according to the different types of batteries in the electrical equipment, which is conducive to improving the accuracy of the health status prediction.

[0173] In some embodiments, Figure 5 This is a flow chart of the detection model training method provided in some other embodiments of the present application. The present application embodiment provides an exemplary introduction to the relevant content of "training the initial detection model according to multiple training samples corresponding to each battery cell type to obtain a preset detection model corresponding to each battery cell type" in the above step S402. Figure 5 As shown, the method of the embodiment of the present application may include the following steps:

[0174] Step S4021: For each cell type, the cell aging training data in the training samples corresponding to the cell type is input into the initial detection model to obtain the training health status output by the initial detection model.

[0175] In this step, for each battery cell type, the detection model training device can input the battery cell aging training data in any training sample corresponding to the battery cell type into the initial detection model to obtain the training health status output by the initial detection model.

[0176] Step S4022: adjust the parameters in the initial detection model according to the cell health status training labels and the training health status in the training samples to obtain an updated initial detection model.

[0177] In this step, the detection model training device can compare the cell health status training label in the training sample with the training health status output by the initial detection model, and adjust the parameters in the initial detection model to obtain an updated initial detection model.

[0178] Step S4023, use other training samples corresponding to the battery cell type as updated training samples, return to execute the step of inputting the battery cell aging training data in the training samples corresponding to the battery cell type into the initial detection model, until the preset training end condition is met, and use the updated initial detection model as the preset detection model corresponding to the battery cell type.

[0179] The preset training end condition in the embodiment of the present application may include but is not limited to any of the following: MSE is less than a preset threshold, or the number of training iterations exceeds a preset threshold.

[0180] Exemplarily, the detection model training device can calculate the MSE by the following formula:

[0181] MSE = avg(||SOH pre -SOH real || 2 )

[0182] Among them, avg() represents the average function; SOH pre Represents the training health status of the initial detection model output; SOH real Represents the battery health status training label.

[0183] Of course, the detection model training device can also calculate the MSE through other modified formulas or equivalent formulas of the above formula.

[0184] In this step, the detection model training device can use other training samples corresponding to the battery cell type as updated training samples, return to execute the above-mentioned step S4021 to input the battery cell aging training data in the training samples corresponding to the battery cell type into the initial detection model, and continue to execute other steps until the preset training end conditions are met, and the updated initial detection model is used as the preset detection model corresponding to the battery cell type.

[0185] In some embodiments, Figure 6 This is a flow chart of the detection model training method provided in some other embodiments of the present application. The present application embodiment provides an exemplary introduction to the relevant content of "inputting the cell aging training data in the training sample corresponding to the cell type into the initial detection model to obtain the training health status output by the initial detection model" in the above step S4021. Figure 6 As shown, the method of the embodiment of the present application may include the following steps:

[0186] Step S601: inputting the cell aging training data in the training samples corresponding to the cell type into the initial detection model.

[0187] In this step, the detection model training device can input the cell aging training data in any training sample corresponding to the cell type into the initial detection model, so that the initial detection model can predict the health status of the cell.

[0188] In some embodiments, the training samples corresponding to any battery cell type in the embodiments of the present application may also include: auxiliary charging condition training data corresponding to the battery cell health status training label. In the embodiments of the present application, the detection model training device inputs the battery cell aging training data and auxiliary charging condition training data in the training samples corresponding to the battery cell type into the initial detection model to facilitate the initial detection model to predict the battery cell health status.

[0189] Exemplarily, the auxiliary charging condition training data of the battery cell can be used to indicate the initial training electrical parameters and / or temperature training parameters of the battery cell in a single training charging process, that is, the auxiliary charging condition training data of the battery cell can be used to indicate different charging training conditions.

[0190] Step S602: extracting a cell aging training feature vector corresponding to the cell aging training data through the initial detection model, and determining a training health state corresponding to the training sample based on the cell aging training feature vector corresponding to the cell aging training data.

[0191] In this step, the detection model training device can extract the cell aging training feature vector corresponding to the cell aging training data in the training sample of the cell type (or simply referred to as the cell aging training feature vector corresponding to the training sample) through the initial detection model, and determine the training health state corresponding to the training sample of the cell type based on the cell aging training feature vector corresponding to the training sample of the cell type.

[0192] In some embodiments, the initial detection model in the embodiments of the present application may include but is not limited to an initial variable-length sequence feature extraction network and an initial battery cell health status prediction network. In the embodiments of the present application, the detection model training device may input the battery cell aging training data in the training samples corresponding to the battery cell type into the initial variable-length sequence feature extraction network to obtain the battery cell aging training feature vector corresponding to the battery cell aging training data.

[0193] In some embodiments, when the cell aging training data and the auxiliary charging condition training data in the training samples corresponding to the cell type are input into the initial detection model in the above step S601, in an embodiment of the present application, the detection model training device can input the cell aging training data and the auxiliary charging condition training data in the training samples corresponding to the cell type into the initial variable-length sequence feature extraction network to obtain the cell aging training feature vector corresponding to the cell aging training data.

[0194] Furthermore, the detection model training device can input the battery cell aging training feature vector into the initial battery cell health state prediction network to obtain the training health state corresponding to the training sample of the battery cell type.

[0195] It should be noted that the implementation method of each step in the embodiment of the present application can refer to the introduction of the relevant steps in the above-mentioned battery cell health status prediction method embodiment. The implementation principles and technical effects are similar and will not be repeated here.

[0196] In some implementations, Figure 7 This is a flowchart of a training sample acquisition method provided in some embodiments of the present application. The present application embodiment provides an exemplary introduction to the relevant content of "obtaining multiple training samples corresponding to each battery cell type" in the above step S401. Figure 7 As shown, the method of the embodiment of the present application may include the following steps:

[0197] Step S701: For each battery cell type, obtain multiple battery cell health status training labels corresponding to the battery cell type.

[0198] In this step, for each battery cell type, the detection model training device can obtain multiple battery cell health status training labels corresponding to the battery cell type.

[0199] In one possible implementation, the detection model training device may determine multiple cell health status training labels corresponding to the cell type based on the historical charging condition data of multiple first reference cells selected corresponding to the cell type. The historical charging condition data may include but is not limited to at least one of the following: an identification code of the historical power-consuming device, a historical cell identification, a charging capacity of the historical cell, a voltage of the historical cell, an initial charging current of the historical cell, an initial charging temperature of the historical cell, a historical charging date, a historical power-on operation time of the power-consuming device, and a historical factory length of the power-consuming device. It should be noted that, in the case where the historical power-consuming device is a vehicle, the power-on operation time of the historical power-consuming device may include but is not limited to information such as historical vehicle mileage.

[0200] In another possible implementation, the detection model training device may obtain multiple battery cell health status training labels corresponding to the battery cell type from other devices.

[0201] Of course, the detection model training device can also obtain multiple battery cell health status training labels corresponding to the battery cell type through other methods.

[0202] Step S702: Determine, according to each battery cell health status training label, a plurality of battery cell aging training data corresponding to each battery cell health status training label.

[0203] In this step, the detection model training device can determine a plurality of battery cell aging training data corresponding to each battery cell health status training label according to each battery cell health status training label.

[0204] Exemplarily, for each battery cell health status training label, the detection model training device can determine multiple battery cell aging training data corresponding to the battery cell health status training label based on the battery cell health status training label. It should be understood that the detection model training device can determine different multiple battery cell aging training data corresponding to the battery cell type based on different battery cell health status training labels.

[0205] In one possible implementation, for each battery cell health status training label, the detection model training device can determine the target battery cell reference parameter corresponding to the first reference battery cell based on the historical charging condition data of the first reference battery cell corresponding to the battery cell health status training label.

[0206] Exemplarily, for each battery cell health status training label, the detection model training device can determine the target battery cell reference parameters corresponding to the first reference battery cell based on information such as the identification code of the historical power-consuming equipment, the historical battery cell identification, the historical power-consuming equipment power-on time and the historical charging date in the historical charging condition data of the first reference battery cell corresponding to the battery cell health status training label; wherein the target battery cell reference parameters may include but are not limited to the identification code of the target power-consuming equipment, the target battery cell identification (i.e., the battery cell identification of the first reference battery cell), the power-on time of the target power-consuming equipment, and the target charging date.

[0207] Furthermore, the detection model training device can obtain other historical charging condition data in which the difference between the cell reference parameters of the first reference cell and the target cell reference parameters meets the preset cell reference conditions, and determine multiple cell aging training data corresponding to the cell health status training label based on the other historical charging condition data.

[0208] Exemplarily, the preset cell reference condition in the embodiment of the present application may mean that the difference in the power-on operation time of the electric device is less than or equal to the preset operation time threshold, and the difference in the charging date of the electric device is less than or equal to the preset charging date threshold.

[0209] It should be noted that the preset operating time threshold and the preset charging date threshold in the embodiment of the present application can be set according to actual conditions, so that the capacity of the battery cell that meets the preset battery cell reference conditions will not be significantly attenuated, so that other historical charging condition data that meet the preset battery cell reference conditions can be consistent with the corresponding battery cell health status training label.

[0210] For example, to facilitate understanding, taking the electric device as a vehicle as an example, the preset battery cell reference condition may mean that the vehicle mileage difference is less than or equal to 1000 km, and the charging date difference is less than or equal to 7 days.

[0211] In the embodiment of the present application, the detection model training device can obtain other historical charging condition data in which the difference between the cell reference parameters of the corresponding cell (i.e., the first reference cell) in the corresponding electrical device and the target cell reference parameters meets the preset cell reference conditions based on the identification code of the target electrical device in the target cell reference parameters and the target cell identifier. Furthermore, the detection model training device can determine multiple cell aging training data corresponding to the cell health status training label based on the historical cell charging data in other historical charging condition data (such as information such as the charging capacity of the historical cell and the voltage of the historical cell).

[0212] It should be noted that the way in which the detection model training device determines the corresponding cell aging training data based on the historical cell charging data can refer to the relevant content of "obtaining the corresponding cell aging data based on the cell charging data" in the above-mentioned cell health status prediction method embodiment, which will not be repeated here.

[0213] Of course, the detection model training device can also determine multiple battery cell aging training data corresponding to the battery cell type in other ways according to the training labels of the health status of each battery cell.

[0214] Step S703: construct multiple training samples corresponding to the cell type according to the health status training labels of the cells and the corresponding multiple cell aging training data.

[0215] In this step, the detection model training device can construct multiple training samples corresponding to the cell type based on the health status training labels of each cell and the corresponding multiple cell aging training data; wherein any training sample can include the health status training label of the cell and the corresponding cell aging training data.

[0216] Exemplarily, for each cell health status training label, the detection model training device can use the cell health status training label and the corresponding cell aging training data as a training sample. It should be understood that the more cell aging training data corresponding to any cell health status training label, the more training samples corresponding to the cell health status training label.

[0217] In summary, in the embodiment of the present application, for each battery cell type, multiple battery cell health status training labels corresponding to the battery cell type are obtained, and multiple battery cell aging training data corresponding to the battery cell type are respectively determined according to each battery cell health status training label. Furthermore, multiple training samples corresponding to the battery cell type are constructed according to each battery cell health status training label and the corresponding multiple battery cell aging training data. It can be seen that in the embodiment of the present application, by determining the corresponding multiple battery cell aging training data according to the battery cell health status training labels, the number of training samples can be accurately expanded in order to improve the accuracy of the detection model training.

[0218] In some implementations, Figure 8 This is a flow chart of a detection model training method provided in some other embodiments of the present application. The present application embodiment provides an exemplary introduction to the relevant content of "obtaining multiple battery cell health status training labels corresponding to the battery cell type" in the above step S701. Figure 8 As shown, the method of the embodiment of the present application may include the following steps:

[0219] Step S7011: screening the historical charging condition data of multiple candidate battery cells corresponding to the battery cell type according to the preset charging voltage condition to obtain the historical charging condition data of multiple first reference battery cells that meet the preset charging voltage condition.

[0220] The preset charging voltage condition in the embodiment of the present application may refer to that the initial charging voltage is less than the first preset voltage threshold cell_v1, and the charging end voltage is greater than the second preset voltage threshold cell_v2, wherein the first preset voltage threshold cell_v1 is less than the second preset voltage threshold cell_v2. The magnitudes of the first preset voltage threshold cell_v1 and the second preset voltage threshold cell_v2 may be set according to the cell type (e.g., ternary cell type or iron-lithium cell type, etc.).

[0221] It should be understood that, considering that only a very small number of vehicles in the market will have such charging conditions (for example, charging conditions when the battery is completely out of power or charging conditions when the battery is fully charged), the first preset voltage threshold cell_v1 does not need to be set to the voltage corresponding to when the battery cell is completely out of power, and the second preset voltage threshold cell_v2 does not need to be set to the voltage corresponding to when the battery cell is fully charged, otherwise, the number of labels will be insufficient.

[0222] In some embodiments, considering that cell aging does not significantly affect the capacity of the high-end and low-end voltage regions (i.e., cell aging causes very little capacity decay in the high-end and low-end voltage regions), the first preset voltage threshold cell_v1 and the second preset voltage threshold cell_v2 may be set in accordance with the principle that the charge capacity in the voltage range below the first preset voltage threshold cell_v1 and above the second preset voltage threshold cell_v2 does not change with cell aging ( Fig. 9 The schematic diagram of the variation curve of the cell voltage as the charging power increases provided in the embodiment of the present application is as follows: Fig. 9 As shown in regions I and II in the figure, the number of labels can be increased by appropriately relaxing the conditions according to the label accuracy requirements).

[0223] In this step, the detection model training device can screen and process the historical charging condition data of multiple candidate battery cells corresponding to the battery cell type according to the preset charging voltage conditions, and use the candidate battery cells whose initial charging voltage and final charging voltage meet the preset charging voltage conditions as the first reference battery cells, so as to obtain the historical charging condition data of the first reference battery cells.

[0224] It should be understood that the same battery cell in the same electrical device may appear more than once in each reference battery cell in the embodiments of the present application, but the corresponding historical power-on operation time or historical charging date must be different.

[0225] Step S7012: for each first reference battery cell, determine the actual capacity of the first reference battery cell according to the historical charging condition data of the first reference battery cell, and determine the battery health status training label of the first reference battery cell according to the ratio of the actual capacity of the first reference battery cell to the rated capacity corresponding to the battery cell type.

[0226] In an embodiment of the present application, the rated capacities corresponding to different battery cell types may be preset in the detection model training device so that the battery cell health status training labels of different battery cell types can be determined.

[0227] In this step, for each first reference battery cell, the detection model training device can determine the actual capacity of the first reference battery cell based on the historical charging condition data of the first reference battery cell, and determine the battery health status training label of the first reference battery cell based on the ratio of the actual capacity of the first reference battery cell to the rated capacity corresponding to the battery type to which the first reference battery cell belongs.

[0228] Exemplarily, the detection model training device may use the ratio of the actual capacity of the first reference battery cell to the rated capacity corresponding to the battery cell type to which the first reference battery cell belongs as the battery cell health status training label of the first reference battery cell.

[0229] As another example, the detection model training device can determine the ratio of the actual capacity of the first reference battery cell to the rated capacity of the battery cell type to which the first reference battery cell belongs; further, the detection model training device can multiply the ratio by a first preset coefficient or increase the first preset coefficient to obtain a corrected ratio, and use the corrected ratio as the battery cell health status training label of the first reference battery cell.

[0230] Of course, the detection model training device can also determine the cell health status training label of the first reference cell in other ways according to the ratio of the actual capacity of the first reference cell to the rated capacity corresponding to the cell type to which the first reference cell belongs.

[0231] In some embodiments, the embodiments of the present application provide an exemplary introduction and explanation of the above-mentioned content of “determining the actual capacity of the first reference battery cell based on the historical charging condition data of the first reference battery cell”.

[0232] Fig.10 A schematic flow chart of a method for determining the actual capacity of a first reference battery cell provided in an embodiment of the present application is shown in FIG. Fig.10 As shown, the method of the embodiment of the present application may include the following steps:

[0233] Step S1001: determining a charging capacity of a first reference battery cell in a first interval of a preset voltage interval according to historical charging condition data of a first reference battery cell.

[0234] In this step, the detection model training device can determine the charging capacity of the first reference cell in the first interval of the preset voltage interval according to the historical charging condition data of the first reference cell. The preset voltage interval is related to the preset charging voltage condition. Exemplarily, the preset voltage interval may include but is not limited to: a voltage interval greater than or equal to the first preset voltage threshold cell_v1 and less than or equal to the second preset voltage threshold cell_v2.

[0235] Exemplarily, the detection model training device can determine the charge capacity cap_v2 of the historical cell corresponding to the voltage of the historical cell equal to the second preset voltage threshold cell_v2, and the charge capacity cap_v1 of the historical cell corresponding to the voltage of the historical cell equal to the first preset voltage threshold cell_v1, based on the charge capacity of the historical cell and the voltage of the historical cell in the historical charging condition data of the first reference cell. Further, the detection model training device can use the difference between the charge capacity cap_v2 and the charge capacity cap_v1 of the historical cell as the first interval charge capacity cap_v12 of the first reference cell in the preset voltage interval.

[0236] Of course, the detection model training device can also determine the charging capacity of the first reference battery cell in the first interval of the preset voltage range in other ways based on the historical charging condition data of the first reference battery cell.

[0237] Step S1002: determining the actual capacity of the first reference battery cell according to the product of the charging capacity in the first interval and the charging capacity correction coefficient corresponding to the battery cell type.

[0238] In an embodiment of the present application, the detection model training device can obtain the charging power correction coefficient corresponding to the battery cell type (i.e., the battery cell type to which the first reference battery cell belongs) from other devices in advance, or can pre-determine the charging power correction coefficient corresponding to the battery cell type (i.e., the battery cell type to which the first reference battery cell belongs).

[0239] In this step, the detection model training device can determine the actual capacity of the first reference battery cell based on the product of the charging capacity cap_v12 in the first interval and the charging capacity correction coefficient corresponding to the battery cell type (i.e., the battery cell type to which the first reference battery cell belongs), so that the battery cell health status training label of the first reference battery cell can be determined based on the actual capacity of the first reference battery cell.

[0240] Exemplarily, the detection model training device may take the product of the first interval charging capacity cap_v12 and the charging capacity correction coefficient f_cap corresponding to the battery cell type as the actual capacity of the first reference battery cell.

[0241] As another example, the detection model training device can determine the product of the charging capacity cap_v12 in the first interval and the charging capacity correction coefficient f_cap corresponding to the battery cell type; further, the detection model training device can increase the above product by a second preset coefficient or multiply it by the second preset coefficient to obtain a corrected product, and use the corrected product as the actual capacity of the first reference battery cell.

[0242] It can be seen that in the implementation of the present application, the charging capacity of the first reference battery cell in the first interval of the preset voltage range is determined according to the historical charging condition data of the first reference battery cell, and the actual capacity of the first reference battery cell is determined according to the product of the charging capacity in the first interval and the charging capacity correction coefficient corresponding to the battery cell type. This realizes the calibration of the battery cell capacity by combining the battery cell aging characteristics, the battery cell voltage and the charging capacity (the operating conditions are relatively stable and quantifiable).

[0243] Of course, the detection model training device can also determine the actual capacity of the first reference battery cell in other ways according to the product of the charging capacity cap_v12 in the first interval and the charging capacity correction coefficient corresponding to the battery cell type.

[0244] In some embodiments, the above-mentioned “method for determining the charging power correction coefficient corresponding to the battery cell type” is exemplarily introduced and explained in the embodiments of the present application.

[0245] Fig.11 A flow chart of a method for determining a charging power correction coefficient provided in an embodiment of the present application is shown in FIG. Fig.11 As shown, the method of the embodiment of the present application may include the following steps:

[0246] Step S1101: screening the historical charging condition data of a plurality of candidate cells corresponding to the cell type according to a preset reference cell condition to obtain the historical charging condition data of a plurality of second reference cells that meet the preset reference cell condition.

[0247] The preset reference cell condition in the embodiment of the present application may mean that the power-on operation time of the electrical equipment is less than the preset power-on operation time threshold, and the factory time of the electrical equipment is less than the preset factory time threshold.

[0248] For example, when the electric device is a vehicle, the startup operation time may include but is not limited to the vehicle mileage. For example, for ease of understanding, taking the electric device as a vehicle as an example, the preset reference cell condition may mean that the vehicle mileage is less than the preset full-charge endurance mileage and the factory time is less than 5 days.

[0249] It should be noted that the preset battery cell usage data threshold and the preset factory length threshold can be adjusted according to actual conditions, as long as the screened second reference battery cells can be close to fresh battery cells without obvious capacity decay.

[0250] In this step, the detection model training device can screen and process the historical charging condition data of multiple candidate battery cells corresponding to the battery cell type according to the preset benchmark battery cell conditions, and use the candidate battery cells whose power-consuming equipment startup time and factory time meet the preset benchmark battery cell conditions as the second reference battery cells, so as to obtain the historical charging condition data of each second reference battery cell.

[0251] Step S1102: determining an average charging capacity of the plurality of second reference cells in a preset voltage range according to historical charging condition data of the plurality of second reference cells.

[0252] In this step, the detection model training device can determine the average charging capacity of the multiple second reference battery cells in the preset voltage range according to the historical charging condition data of the multiple second reference battery cells.

[0253] Exemplarily, the detection model training device can determine the second interval charging capacity of each second reference battery cell in a preset voltage range based on the historical charging condition data of the multiple second reference battery cells, and determine the average interval charging capacity of the multiple second reference battery cells in the preset voltage range based on the second interval charging capacity of each second reference battery cell in the preset voltage range.

[0254] In a possible implementation, for each second reference battery cell, the detection model training device may determine the second interval charging capacity of the second reference battery cell in a preset voltage range based on the historical charging condition data of the second reference battery cell.

[0255] Exemplarily, for each second reference cell, the detection model training device can determine the historical cell charge capacity cap_v2' of the second reference cell corresponding to the historical cell voltage equal to the second preset voltage threshold cell_v2, and the historical cell charge capacity cap_v1' of the second reference cell corresponding to the historical cell voltage equal to the first preset voltage threshold cell_v1, based on the historical cell charge capacity and the historical cell voltage in the historical charging condition data of the second reference cell. Further, the detection model training device can use the difference between the charge capacity cap_v2' and the charge capacity cap_v1' of the historical cell as the second interval charge capacity cap_v12' of the second reference cell in the preset voltage interval.

[0256] Furthermore, the detection model training device can calculate the average value of the second interval charging capacity of each second reference battery cell to obtain the interval charging capacity average value.

[0257] Of course, the detection model training device can also determine the average charging capacity of multiple second reference batteries in a preset voltage range in other ways based on the historical charging condition data of multiple second reference batteries.

[0258] Step S1103: Determine a charge power correction coefficient corresponding to the cell type according to a ratio of an average charge power value in the interval to a rated capacity corresponding to the cell type.

[0259] In this step, the detection model training device can determine the charging power correction coefficient corresponding to the cell type according to the ratio of the interval charging power average value to the rated capacity corresponding to the cell type (i.e., the cell type to which the second reference cell belongs), so that the real capacity of each reference cell can be determined according to the product of the charging power correction coefficient corresponding to the cell type. It should be understood that the first reference cell and the second reference cell in the embodiment of the present application belong to the same cell type.

[0260] Exemplarily, the detection model training device may use the ratio of the interval charging capacity average value to the rated capacity corresponding to the battery cell type as the charging capacity correction coefficient corresponding to the battery cell type.

[0261] As another example, the detection model training device can determine the ratio of the average charging capacity of the interval to the rated capacity corresponding to the battery cell type; further, the detection model training device can increase the above ratio by a third preset coefficient or multiply it by the third preset coefficient to obtain a correction ratio, and use the correction ratio as the charging capacity correction coefficient corresponding to the battery cell type.

[0262] Of course, the detection model training device can also determine the charging capacity correction coefficient corresponding to the battery cell type in other ways based on the ratio of the average charging capacity of the interval to the rated capacity corresponding to the battery cell type.

[0263] It should be understood that in the embodiments of the present application, the charging power correction coefficients corresponding to different battery cell types need to be formulated separately, which may be affected by the type of battery cell and / or the size of the battery cell capacity.

[0264] In summary, in the embodiment of the present application, the historical charging condition data of multiple candidate cells corresponding to the cell type are screened according to the preset charging voltage condition to obtain the historical charging condition data of multiple first reference cells that meet the preset charging voltage condition. Further, for each first reference cell, the real capacity of the first reference cell is determined according to the historical charging condition data of the first reference cell, and the cell health status training label of the first reference cell is determined according to the ratio of the real capacity of the first reference cell to the rated capacity corresponding to the cell type. It can be seen that in the embodiment of the present application, by screening out the historical charging condition data of the first reference battery cell according to the preset charging voltage conditions, and determining the actual capacity of the first reference battery cell, and further, determining the method of training the battery cell health status according to the ratio of the actual capacity of the first reference battery cell to the rated capacity, the charging condition data of the battery cell based on market power-consuming equipment is realized without the aid of other data indicators generated by dependent algorithms (which can avoid the influence of other errors), and calibrating the battery cell health status training labels in combination with the battery cell aging characteristics and voltage limit conditions. This can not only provide real and accurate battery cell health status training labels, laying the foundation for improving the performance of the detection model, but also can be applied to different charging conditions, and has strong practicality.

[0265] In some embodiments, based on the above embodiments, the overall process of the detection model training method and the battery cell health status prediction method is exemplarily introduced and explained in the embodiments of the present application by taking the electric equipment as a vehicle as an example. Fig.12 The overall flow chart of the detection model training method and the battery cell health status prediction method provided in the embodiments of the present application is provided. The present application embodiment takes the method applied to a computer device as an example for explanation, wherein the computer device may include but is not limited to a battery cell health status prediction device or a detection model training device. Fig.12 As shown, the method of the embodiment of the present application may include the following steps:

[0266] Step S1201: The computer device obtains historical charging condition data of battery cells of different battery cell types of different vehicles during historical single charging processes from a power battery big data platform.

[0267] It should be noted that in the embodiments of the present application, battery cells from different battery cell manufacturers and historical charging condition data under different charging conditions will be obtained, so that the trained preset detection model can be applicable to the health status prediction of various battery cell types and under various charging conditions.

[0268] Step S1202: For each cell type, the computer device screens historical charging condition data of multiple candidate cells corresponding to the cell type according to a preset charging voltage condition to obtain historical charging condition data of multiple first reference cells that meet the preset charging voltage condition.

[0269] It should be understood that in the above step S1201, the computer device obtains the historical charging condition data of the battery cell of any battery cell type, wherein the battery cell corresponding to any historical charging condition data can be referred to as a candidate battery cell.

[0270] Step S1203: For each first reference battery cell of each battery cell type, the computer device determines a first interval charging capacity of the first reference battery cell in a preset voltage interval according to historical charging condition data of the first reference battery cell.

[0271] Step S1204: For each battery cell type, the computer device determines a charge power correction coefficient corresponding to the battery cell type.

[0272] Step S1205: For each first reference battery cell of each battery cell type, the computer device determines a battery health status training label of the first reference battery cell according to the first interval charging capacity of the first reference battery cell in the preset voltage interval and the charging capacity correction coefficient corresponding to the battery cell type.

[0273] It should be understood that the above steps S1201 to S1205 complete the calibration process of the battery cell health status training label.

[0274] Step S1206: For each battery cell health status training label of each battery cell type, the computer device obtains a plurality of battery cell aging training data corresponding to the battery cell health status training label from the power battery big data platform according to the battery cell health status training label.

[0275] It should be understood that the computer device can obtain multiple other historical charging condition data corresponding to the battery health status training label from the power battery big data platform based on the battery health status training label; further, the computer device can determine the multiple battery cell aging training data corresponding to the battery health status training label based on information such as the historical battery cell charging capacity and the historical battery cell voltage in other historical charging condition data.

[0276] Step S1207: For each cell type, the computer device constructs multiple training samples corresponding to the cell type according to the health status training labels of each cell and the corresponding multiple cell aging training data, that is, constructs a training data set.

[0277] Step S1208: The computer device constructs a preset detection model corresponding to each battery cell type according to the training data set.

[0278] Exemplarily, the preset detection model corresponding to any battery cell type may include, but is not limited to: a variable-length sequence feature extraction network and a battery cell health status prediction network.

[0279] Step S1209: The computer device may deploy the preset detection model corresponding to each battery cell type in the cloud.

[0280] Step S1210: The computer device may call a preset detection model corresponding to any battery cell type to predict the health status of the battery cell.

[0281] Exemplarily, when a computer device obtains the charging condition data of a single charging process of any electrical device, it determines the battery type and battery aging data of each battery cell in the electrical device based on the charging condition data, and calls the preset detection model corresponding to the corresponding battery cell type to predict the health status of the battery cell.

[0282] It should be noted that the implementable methods of each step in the embodiments of the present application can refer to the relevant contents in the above embodiments of the present application, which will not be repeated here.

[0283] In summary, the implementation of this application provides a calibration method, a detection model training method and a cell-level health status prediction method for training labels for battery health status based entirely on big data of market power equipment. The entire prediction method is completely based on market power equipment data without the help of any laboratory test data. Different from the aging of batteries under a relatively single charging condition in laboratory testing, the aging of batteries in market power equipment is the capacity attenuation under the combined influence of various practical factors such as different temperatures, static conditions, charging behaviors and usage habits (such as driving habits). There are inevitable differences in the aging mechanisms of batteries in the two forms. Therefore, the embodiment of the present application can provide real and accurate battery health status training labels for different battery types under different charging conditions, so that the preset detection model obtained by training is more accurate, and the prediction method based on the preset detection model can more accurately capture the battery cell aging characteristics in the power equipment in the market, and has higher prediction accuracy in practical applications. In addition, the embodiment of the present application realizes the prediction of battery cell-level health status, rather than simply predicting the health status of the entire battery pack. The prediction of battery cell-level health status can provide more valuable information for abnormal battery cell detection, after-sales maintenance, etc.

[0284] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0285] Based on the same inventive concept, the embodiment of the present application also provides a battery cell health status prediction device for implementing the battery cell health status prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more battery cell health status prediction device embodiments provided below can refer to the limitations of the battery cell health status prediction method above, and will not be repeated here.

[0286] In some embodiments, Fig.13 This is a schematic diagram of the structure of the battery cell health status prediction device in some embodiments of the present application. The battery cell health status prediction device provided in the embodiments of the present application can be applied to the battery cell health status prediction device. Fig.13 As shown, the battery cell health status prediction device of the embodiment of the present application may include: an acquisition module 1301 , a first determination module 1302 , an input module 1303 and a prediction module 1304 .

[0287] The acquisition module 1301 is used to acquire charging condition data of a single charging process of an electric device;

[0288] A first determination module 1302 is used to determine the cell type and cell aging data of each cell in the electric device according to the charging condition data;

[0289] An input module 1303, used to input the cell aging data of each cell into a preset detection model corresponding to the cell type;

[0290] The prediction module 1304 is used to extract the cell aging feature vector of each cell through a preset detection model, and determine the predicted health state of each cell based on the cell aging feature vector of each cell.

[0291] In some embodiments, the preset detection model includes a variable length sequence feature extraction network and a cell health status prediction network, and the prediction module 1304 includes:

[0292] An extraction unit, used to input the cell aging data of each cell into a variable-length sequence feature extraction network to obtain a cell aging feature vector of each cell;

[0293] The prediction unit is used to input the cell aging feature vector of each cell into the cell health state prediction network to obtain the predicted health state of each cell.

[0294] In some embodiments, the input module 1303 is specifically used to:

[0295] Inputting the cell aging data and auxiliary charging condition data of each cell into a preset detection model corresponding to the cell type respectively; wherein the auxiliary charging condition data of the cell is used to indicate the initial electrical parameters and / or temperature parameters of the cell in a single charging process;

[0296] Correspondingly, the extraction unit is specifically used for:

[0297] For each battery cell, the battery cell aging data and auxiliary charging condition data of the battery cell are input into a variable-length sequence feature extraction network to obtain a battery cell aging feature vector of the battery cell.

[0298] In some embodiments, the first determining module 1302 is specifically configured to:

[0299] Determine the cell type and cell charging data of each cell in the electric device according to the charging condition data;

[0300] Data preprocessing is performed on the battery cell charging data of each battery cell to obtain the battery cell aging data of each battery cell.

[0301] In some embodiments, the battery cell health status prediction device further includes:

[0302] The second determination module is used to determine the auxiliary charging condition data of each battery cell according to the charging condition data.

[0303] The battery cell health status prediction device provided in the embodiment of the present application can be used to execute the technical solution in the above-mentioned battery cell health status prediction method embodiment of the present application. Its implementation principle and technical effect are similar and will not be repeated here.

[0304] Each module in the above-mentioned battery cell health status prediction device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the battery cell health status prediction device in the form of hardware, or can be stored in the memory in the battery cell health status prediction device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0305] Based on the same inventive concept, the embodiment of the present application also provides a detection model training device for implementing the above-mentioned battery cell health status prediction method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more detection model training device embodiments provided below can refer to the limitations of the detection model training method above, and will not be repeated here.

[0306] In some embodiments, Fig.14 This is a schematic diagram of the structure of the detection model training device in some embodiments of the present application. The detection model training device provided in the embodiments of the present application can be applied to the detection model training device. Fig.14 As shown, the detection model training device of the embodiment of the present application may include: an acquisition module 1401 and a training module 1402.

[0307] The acquisition module 1401 is used to acquire multiple training samples corresponding to each battery cell type; wherein the training samples corresponding to the battery cell type include: battery cell health status training labels and corresponding battery cell aging training data;

[0308] The training module 1402 is used to train the initial detection model according to a plurality of training samples corresponding to each battery cell type, so as to obtain a preset detection model corresponding to each battery cell type.

[0309] In some embodiments, the training module 1402 includes:

[0310] A prediction unit, for each cell type, inputting the cell aging training data in the training samples corresponding to the cell type into the initial detection model to obtain the training health status output by the initial detection model;

[0311] An adjustment unit, used for adjusting the parameters in the initial detection model according to the cell health status training labels and the training health status in the training samples to obtain an updated initial detection model;

[0312] The loop unit is used to use other training samples corresponding to the battery cell type as updated training samples, return to execute the step of inputting the battery cell aging training data in the training samples corresponding to the battery cell type into the initial detection model, until the preset training end condition is met, and the updated initial detection model is used as the preset detection model corresponding to the battery cell type.

[0313] In some embodiments, the prediction unit includes:

[0314] An input subunit, used for inputting the cell aging training data in the training samples corresponding to the cell type into the initial detection model;

[0315] The prediction subunit is used to extract the cell aging training feature vector corresponding to the cell aging training data through the initial detection model, and determine the training health state corresponding to the training sample based on the cell aging training feature vector corresponding to the cell aging training data.

[0316] In some embodiments, the initial detection model includes an initial variable length sequence feature extraction network and an initial cell health status prediction network, and the prediction subunit is specifically used for:

[0317] Inputting the cell aging training data in the training samples corresponding to the cell type into the initial variable-length sequence feature extraction network to obtain the cell aging training feature vector corresponding to the cell aging training data;

[0318] The cell aging training feature vector is input into the initial cell health state prediction network to obtain the training health state corresponding to the training sample.

[0319] In some embodiments, the training samples corresponding to the cell type further include: auxiliary charging condition training data corresponding to the cell health status training label, and the input subunit is specifically used for:

[0320] Inputting the cell aging training data and auxiliary charging condition training data in the training samples corresponding to the cell type into the initial detection model;

[0321] Correspondingly, the prediction subunit is specifically used for:

[0322] The cell aging training data and the auxiliary charging condition training data are input into the initial variable-length sequence feature extraction network to obtain the cell aging training feature vector corresponding to the cell aging training data.

[0323] In some embodiments, the acquisition module 1401 includes:

[0324] An acquisition unit, used for acquiring, for each battery cell type, a plurality of battery cell health status training labels corresponding to the battery cell type;

[0325] A first determining unit, configured to determine, according to each battery cell health status training label, a plurality of battery cell aging training data corresponding to each battery cell health status training label;

[0326] The construction unit is used to construct multiple training samples corresponding to the battery cell type according to the health status training labels of each battery cell and the corresponding multiple battery cell aging training data.

[0327] In some embodiments, the acquisition unit includes:

[0328] A screening subunit, configured to screen the historical charging condition data of a plurality of candidate battery cells corresponding to the battery cell type according to a preset charging voltage condition, and obtain the historical charging condition data of a plurality of first reference battery cells that meet the preset charging voltage condition;

[0329] A determination subunit is used to determine, for each first reference battery cell, the actual capacity of the first reference battery cell according to the historical charging condition data of the first reference battery cell, and to determine the battery cell health status training label of the first reference battery cell according to the ratio of the actual capacity of the first reference battery cell to the rated capacity corresponding to the battery cell type.

[0330] In some embodiments, the determining subunit is specifically configured to:

[0331] Determine the charging capacity of the first reference battery cell in a first interval of a preset voltage range according to the historical charging condition data of the first reference battery cell;

[0332] The actual capacity of the first reference battery cell is determined according to the product of the charging capacity in the first interval and the charging capacity correction coefficient corresponding to the battery cell type.

[0333] In some embodiments, the acquisition module 1401 further includes:

[0334] A screening unit, configured to screen the historical charging condition data of a plurality of candidate cells corresponding to the cell type according to a preset reference cell condition, and obtain the historical charging condition data of a plurality of second reference cells that meet the preset reference cell condition;

[0335] A second determining unit, configured to determine an average charging capacity of the plurality of second reference cells in a preset voltage range according to historical charging condition data of the plurality of second reference cells;

[0336] The third determining unit is used to determine the charging power correction coefficient corresponding to the battery cell type according to the ratio of the interval charging power average value to the rated capacity corresponding to the battery cell type.

[0337] In some embodiments, the second determining unit is specifically configured to:

[0338] For each second reference battery cell, determining a second interval charging capacity of the second reference battery cell in a preset voltage range according to historical charging condition data of the second reference battery cell;

[0339] The average value of the second interval charging power of each second reference battery cell is calculated to obtain the interval charging power average value.

[0340] In some embodiments, the first determining unit is specifically configured to:

[0341] For each battery cell health status training label, determine a target battery cell reference parameter corresponding to the first reference battery cell according to the historical charging condition data of the first reference battery cell corresponding to the battery cell health status training label;

[0342] Obtain other historical charging condition data in which the difference between the cell reference parameter of the first reference cell and the target cell reference parameter meets the preset cell reference condition, and determine multiple cell aging training data corresponding to the cell health status training label based on the other historical charging condition data.

[0343] The detection model training device provided in the embodiment of the present application can be used to execute the technical solution in the above-mentioned detection model training method embodiment of the present application. Its implementation principle and technical effect are similar and will not be repeated here.

[0344] Each module in the above detection model training device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above modules can be embedded in or independent of the processor in the detection model training device in the form of hardware, or can be stored in the memory in the detection model training device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0345] In some embodiments, Fig.15 Schematic diagram of the structure of a computer device in some embodiments of the present application. For example, the computer device in the embodiments of the present application may include but is not limited to a battery cell health status prediction device or a detection model training device. Fig.15 As shown, the computer device provided in the embodiment of the present application may include a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store relevant data or information in the battery health status prediction process and / or the detection model training process. The network interface of the computer device is used to communicate with an external device through a network connection. When the computer program is executed by the processor to implement the technical solution in the above-mentioned battery health status prediction method embodiment or the above-mentioned detection model training method embodiment of the present application, its implementation principle and technical effect are similar, and will not be repeated here.

[0346] Those skilled in the art will understand that Fig.15 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0347] In some embodiments, a battery cell health status prediction device is also provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the technical solution in the above-mentioned battery cell health status prediction method embodiment of the present application is implemented. The implementation principle and technical effect are similar and will not be repeated here.

[0348] In some embodiments, a detection model training device is also provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the technical solution in the above-mentioned detection model training method embodiment of the present application is implemented. The implementation principle and technical effect are similar and will not be repeated here.

[0349] In some embodiments, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the technical solution of the above-mentioned battery cell health status prediction method embodiment or the above-mentioned detection model training method embodiment of the present application is implemented. The implementation principle and technical effects are similar and will not be repeated here.

[0350] In some embodiments, a computer program product is provided, including a computer program. When the computer program is executed by a processor, it implements the technical solution in the above-mentioned battery cell health status prediction method embodiment or the above-mentioned detection model training method embodiment of the present application. The implementation principle and technical effects are similar and will not be repeated here.

[0351] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The processor involved in each embodiment provided in this application can be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, etc., but is not limited thereto.

[0352] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application. In particular, as long as there is no structural conflict, the various technical features mentioned in the various embodiments can be combined in any way. The present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.

Claims

1. A method for predicting the health status of a battery cell, characterized in that: The method comprises: Obtain charging condition data of a single charging process of an electric device; Determine the cell type and cell aging data of each cell in the electrical device according to the charging condition data; Inputting the cell aging data of each of the cell into a preset detection model corresponding to the cell type; The cell aging feature vectors of each of the cell are extracted respectively through the preset detection model, and the predicted health status of each of the cell is determined based on the cell aging feature vectors of each of the cell.

2. The method according to claim 1, characterized in that The preset detection model includes a variable-length sequence feature extraction network and a battery cell health state prediction network. The battery cell aging feature vectors of each battery cell are extracted respectively through the preset detection model, and the predicted health state of each battery cell is determined based on the battery cell aging feature vectors of each battery cell, including: Inputting the cell aging data of each of the cell into the variable-length sequence feature extraction network respectively to obtain a cell aging feature vector of each of the cell; The cell aging feature vectors of the battery cells are respectively input into the battery cell health state prediction network to obtain the predicted health state of the battery cells.

3. The method according to claim 2, characterized in that The step of inputting the cell aging data of each of the cells into a preset detection model corresponding to the cell type comprises: Inputting the cell aging data and auxiliary charging condition data of each cell into a preset detection model corresponding to the cell type respectively; wherein the auxiliary charging condition data of the cell is used to indicate the initial electrical parameters and / or temperature parameters of the cell during the single charging process; Correspondingly, the cell aging data of each of the cells is respectively input into the variable-length sequence feature extraction network to obtain a cell aging feature vector of each of the cells, including: For each of the battery cells, the battery cell aging data and the auxiliary charging condition data of the battery cell are input into the variable-length sequence feature extraction network to obtain the battery cell aging feature vector of the battery cell.

4. The method according to any one of claims 1 to 3, characterized in that The determining the cell type and cell aging data of each cell in the electric device according to the charging condition data includes: Determine the cell type and cell charging data of each cell in the electrical device according to the charging condition data; Data preprocessing is performed on the battery cell charging data of each of the battery cells to obtain battery cell aging data of each of the battery cells.

5. The method according to claim 4, characterized in that The method further comprises: The auxiliary charging condition data of each of the battery cells is determined according to the charging condition data.

6. A detection model training method, characterized in that: The method comprises: Acquire multiple training samples corresponding to each battery cell type; wherein the training samples corresponding to the battery cell type include: battery cell health status training labels and corresponding battery cell aging training data; The initial detection model is trained according to a plurality of training samples corresponding to each of the battery cell types to obtain a preset detection model corresponding to each of the battery cell types.

7. The method according to claim 6, characterized in that The initial detection model is trained according to a plurality of training samples corresponding to each of the battery cell types to obtain a preset detection model corresponding to each of the battery cell types, including: For each of the battery cell types, inputting battery cell aging training data in the training samples corresponding to the battery cell type into the initial detection model to obtain a training health state output by the initial detection model; Adjust the parameters in the initial detection model according to the cell health status training labels in the training samples and the training health status to obtain an updated initial detection model; Use other training samples corresponding to the battery cell type as updated training samples, return to execute the step of inputting the battery cell aging training data in the training samples corresponding to the battery cell type into the initial detection model, until the preset training end condition is met, and use the updated initial detection model as the preset detection model corresponding to the battery cell type.

8. The method according to claim 7, characterized in that The step of inputting the cell aging training data in the training samples corresponding to the cell type into the initial detection model to obtain the training health status output by the initial detection model includes: Inputting the cell aging training data in the training samples corresponding to the cell type into the initial detection model; A cell aging training feature vector corresponding to the cell aging training data is extracted through the initial detection model, and a training health state corresponding to the training sample is determined based on the cell aging training feature vector corresponding to the cell aging training data.

9. The method according to claim 8, characterized in that The initial detection model includes an initial variable-length sequence feature extraction network and an initial cell health state prediction network. The initial detection model is used to extract a cell aging training feature vector corresponding to the cell aging training data, and the training health state corresponding to the training sample is determined based on the cell aging training feature vector corresponding to the cell aging training data, including: Inputting the cell aging training data in the training samples corresponding to the cell type into the initial variable-length sequence feature extraction network to obtain a cell aging training feature vector corresponding to the cell aging training data; The cell aging training feature vector is input into the initial cell health state prediction network to obtain the training health state corresponding to the training sample.

10. The method according to claim 8, characterized in that The training samples corresponding to the cell type also include: auxiliary charging condition training data corresponding to the cell health status training label, and the cell aging training data in the training samples corresponding to the cell type is input into the initial detection model, including: Inputting the battery cell aging training data and the auxiliary charging condition training data in the training samples corresponding to the battery cell type into the initial detection model; Correspondingly, the step of inputting the cell aging training data in the training samples corresponding to the cell type into the initial variable-length sequence feature extraction network to obtain the cell aging training feature vector corresponding to the cell aging training data includes: The cell aging training data and the auxiliary charging condition training data are input into the initial variable-length sequence feature extraction network to obtain a cell aging training feature vector corresponding to the cell aging training data.

11. The method according to any one of claims 6 to 10, characterized in that: The obtaining of multiple training samples corresponding to each battery cell type includes: For each of the battery cell types, obtaining a plurality of battery cell health status training labels corresponding to the battery cell type; Determine, according to each of the battery cell health status training labels, a plurality of battery cell aging training data corresponding to each of the battery cell health status training labels; According to each of the battery cell health status training labels and the corresponding multiple battery cell aging training data, multiple training samples corresponding to the battery cell type are constructed.

12. The method according to claim 11, characterized in that The obtaining of a plurality of battery cell health status training labels corresponding to the battery cell type includes: Screening and processing the historical charging condition data of multiple candidate battery cells corresponding to the battery cell type according to a preset charging voltage condition to obtain the historical charging condition data of multiple first reference battery cells that meet the preset charging voltage condition; For each of the first reference battery cells, the actual capacity of the first reference battery cell is determined based on the historical charging condition data of the first reference battery cell, and the battery health status training label of the first reference battery cell is determined based on the ratio of the actual capacity of the first reference battery cell to the rated capacity corresponding to the battery cell type.

13. The method according to claim 12, characterized in that The determining the actual capacity of the first reference battery cell according to the historical charging condition data of the first reference battery cell comprises: Determining a charging capacity of the first reference battery cell in a first interval of a preset voltage range according to historical charging condition data of the first reference battery cell; The actual capacity of the first reference battery cell is determined according to the product of the charging capacity in the first interval and the charging capacity correction coefficient corresponding to the battery cell type.

14. The method according to claim 13, characterized in that The method further comprises: Screening and processing the historical charging condition data of multiple candidate cells corresponding to the cell type according to a preset reference cell condition to obtain historical charging condition data of multiple second reference cells that meet the preset reference cell condition; Determine an average charging capacity of the plurality of second reference cells in the preset voltage range according to the historical charging condition data of the plurality of second reference cells; The charging power correction coefficient corresponding to the battery cell type is determined according to the ratio of the average charging power value of the interval to the rated capacity corresponding to the battery cell type.

15. The method according to claim 14, characterized in that The step of determining the average charging capacity of the plurality of second reference cells in the preset voltage range according to the historical charging condition data of the plurality of second reference cells includes: For each of the second reference cells, determining a charging capacity of the second reference cell in a second interval of the preset voltage interval according to the historical charging condition data of the second reference cell; The average value of the second interval charging power of each of the second reference batteries is calculated to obtain the interval charging power average value.

16. The method according to claim 12, characterized in that The determining, according to each of the battery cell health status training labels, a plurality of battery cell aging training data corresponding to each of the battery cell health status training labels respectively includes: For each of the battery cell health status training labels, determining a target battery cell reference parameter corresponding to the first reference battery cell according to the historical charging condition data of the first reference battery cell corresponding to the battery cell health status training label; Obtain other historical charging condition data in which the difference between the cell reference parameter of the first reference cell and the target cell reference parameter meets the preset cell reference condition, and determine multiple cell aging training data corresponding to the cell health status training label based on the other historical charging condition data.

17. A battery cell health status prediction device, characterized in that: The device comprises: An acquisition module is used to acquire charging condition data of a single charging process of an electric device; A determination module, used to determine the cell type and cell aging data of each cell in the electrical device according to the charging condition data; An input module, used for inputting the cell aging data of each of the cells into a preset detection model corresponding to the cell type; The prediction module is used to extract the cell aging feature vectors of each of the battery cells respectively through the preset detection model, and determine the predicted health status of each of the battery cells based on the cell aging feature vectors of each of the battery cells respectively.

18. A detection model training device, characterized in that: The device comprises: An acquisition module is used to acquire multiple training samples corresponding to each battery cell type; wherein the training samples corresponding to the battery cell type include: battery cell health status training labels and corresponding battery cell aging training data; The training module is used to train the initial detection model according to multiple training samples corresponding to each of the battery cell types to obtain a preset detection model corresponding to each of the battery cell types.

19. A battery cell health status prediction device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

20. A detection model training device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 6 to 16 are implemented.

21. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 16 are implemented.

22. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 16 are implemented.

Images