A battery state of health assessment method, system, device, and medium

Abstract

The application discloses a battery health state evaluation method, system, device and medium, wherein the method comprises: performing energy feature extraction processing on a target battery to obtain sequence energy features; performing data conversion processing on the sequence energy features to obtain multi-dimensional energy features; inputting the multi-dimensional energy features into a state evaluation model to perform battery health state evaluation processing to obtain the health state of the target battery, and the state evaluation model is a hybrid model based on the fusion of deep learning and machine learning. Through the extraction of the energy features and the battery health state evaluation based on the state evaluation model of the fusion of deep learning and machine learning, the accuracy of the evaluation result can be improved, and the application can be widely applied to the technical field of battery management.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and in particular to a method, system, device and medium for assessing battery health status. Background Technology

[0002] Lithium-ion batteries possess advantages such as high energy density, long lifespan, fast charging speed, and environmental friendliness, and are widely used in electric vehicles, mobile phones, laptops, and other fields. However, battery health deteriorates with usage time, making accurate estimation of battery health crucial. Current methods for estimating battery health primarily involve analyzing simple battery features input into a model. However, these simple features cannot fully represent the comprehensive information about battery degradation, resulting in inaccurate assessments. Therefore, the technical problems existing in these technologies urgently need to be addressed. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a battery health status assessment method, system, device, and medium to improve the accuracy of battery health status assessment.

[0004] On one hand, the present invention provides a method for assessing battery health status, comprising:

[0005] Energy feature extraction is performed on the target battery to obtain sequential energy features;

[0006] The sequence energy features are subjected to data transformation processing to obtain multidimensional energy features;

[0007] The multidimensional energy characteristics are input into the state assessment model to perform battery health state assessment processing, thereby obtaining the health state of the target battery. The state assessment model is a hybrid model based on the fusion of deep learning and machine learning.

[0008] Optionally, the step of extracting energy features from the target battery to obtain sequential energy features includes:

[0009] Determine the sampling range of the target battery;

[0010] The target battery is sampled at equal intervals according to the sampling interval to obtain voltage data and current data.

[0011] The voltage and current data are linearly interpolated according to the energy characteristic calculation formula to obtain the sequence energy characteristics.

[0012] Optionally, the step of performing linear interpolation on the voltage data and the current data according to the energy characteristic calculation formula to obtain the sequence energy characteristics includes:

[0013] Obtain the sampling time, sampling voltage, and sampling current corresponding to each sampling point of the linear interpolation;

[0014] The energy characteristics of the sequence are calculated by multiplying the sampling voltage and the sampling current according to the energy characteristic calculation formula and integrating them with the sampling time.

[0015] Optionally, the step of performing data transformation processing on the sequence energy features to obtain multidimensional energy features includes:

[0016] A one-dimensional energy characteristic curve is constructed based on the sequence energy characteristics. The one-dimensional energy characteristic curve is a curve with voltage as the horizontal axis and energy as the vertical axis.

[0017] The one-dimensional energy characteristic curve is sampled using a sliding window to obtain multi-dimensional energy characteristics.

[0018] Optionally, the step of inputting the multidimensional energy characteristics into the state assessment model for battery health state assessment to obtain the health state of the target battery includes:

[0019] Input the multidimensional energy characteristics into the state evaluation model;

[0020] The first module of the state assessment model performs feature extraction processing on the multidimensional energy features to obtain the semantic information of the multidimensional energy features;

[0021] The health status of the target battery is obtained by performing regression prediction processing on the semantic information through the second module of the state assessment model.

[0022] Optionally, before inputting the multidimensional energy features into the state assessment model for battery health state assessment, the method further includes pre-training the state assessment model, specifically including:

[0023] Build a state assessment model based on deep learning and machine learning;

[0024] By selecting multiple training batteries from a large battery dataset;

[0025] The training battery is subjected to multidimensional energy feature extraction processing to obtain a training dataset;

[0026] The training dataset is input into the state evaluation model, and the state evaluation model is trained by cross-validation.

[0027] Optionally, the step of constructing a state evaluation model based on deep learning and machine learning includes:

[0028] Obtain a deep learning-based residual network model and a machine learning-based support vector machine model;

[0029] The fully connected layers of the residual network model are replaced according to the support vector machine model to obtain the state evaluation model.

[0030] On the other hand, embodiments of the present invention also provide a battery health status assessment system, including:

[0031] The first module is used to extract energy features from the target battery to obtain sequential energy features;

[0032] The second module is used to perform data conversion processing on the sequence energy features to obtain multidimensional energy features;

[0033] The third module is used to input the multidimensional energy characteristics into the state assessment model for battery health state assessment to obtain the health state of the target battery. The state assessment model is a hybrid model based on the fusion of deep learning and machine learning.

[0034] On the other hand, embodiments of the present invention also disclose an electronic device, including a processor and a memory;

[0035] The memory is used to store programs;

[0036] The processor executes the program to implement the method described above.

[0037] On the other hand, embodiments of the present invention also disclose a computer-readable storage medium storing a program that is executed by a processor to implement the methods described above.

[0038] On the other hand, embodiments of the present invention also disclose a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned method.

[0039] Compared with existing technologies, the present invention, employing the above technical solutions, has the following technical effects: The embodiments of the present invention perform energy feature extraction processing on the target battery to obtain sequential energy features, which can better capture the voltage and current changes during battery aging through time-series-based energy features; furthermore, the embodiments of the present invention perform data transformation processing on the sequential energy features to obtain multi-dimensional energy features, which can better mine the changes in data in time and space, and is more conducive to training the state assessment model; moreover, the state assessment model of the embodiments of the present invention is a hybrid model based on the fusion of deep learning and machine learning, which can better assess the battery health status, thereby improving the accuracy of battery health status assessment. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart of a battery health status assessment method provided in an embodiment of this application;

[0042] Figure 2 This is a schematic diagram of the energy-time curve during battery charging provided in an embodiment of this application;

[0043] Figure 3 This is a schematic diagram of the energy-voltage curve during battery charging provided in an embodiment of this application;

[0044] Figure 4 This is a schematic diagram of a two-dimensional energy feature generation process provided in an embodiment of this application;

[0045] Figure 5 This is a schematic diagram of a sliding window provided in an embodiment of this application;

[0046] Figure 6 This is a schematic diagram showing the comparison between the estimated health status of a residual network and the actual health status provided in an embodiment of this application.

[0047] Figure 7 This is a schematic diagram showing the comparison between the estimated health status value and the actual value of a health status assessment model provided in this application embodiment;

[0048] Figure 8 This is a schematic diagram of the error result of a residual network provided in an embodiment of this application;

[0049] Figure 9 This is a schematic diagram of the error results of a state assessment model provided in an embodiment of this application. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] First, let's analyze some of the terms used in this application:

[0052] Lithium-ion battery health characteristics: one or more independent, observable properties or characteristics of a lithium-ion battery that can be reflected by simplified numerical concepts, such as the battery's state of charge and capacity.

[0053] Deep learning is a machine learning method based on learning representations of data. Its unique topological structure enables the active extraction of features, learning low-level and high-level features from data, and combining feature extractors and classifiers to provide end-to-end solutions to problems.

[0054] ResNet: Residual Neural Network, is a type of convolutional neural network that offers advantages over conventional convolutional neural networks. Generally, convolutional neural networks degenerate as their depth increases. ResNet addresses this issue of poor training performance when networks are too deep by utilizing a residual structure.

[0055] SVR: Support Vector Regression, is a machine learning algorithm used to analyze regression problems in data. The input vector is mapped to a high-dimensional space, and a hyperplane is defined in this space. For regression problems, this hyperplane is adjusted to minimize the distance to the farthest sample point in the data, thus transforming a nonlinear problem into a linear problem in a high-dimensional space by fitting the hyperplane to the data.

[0056] Current methods for estimating battery health status mainly include model-based methods and data-driven methods. Model-based methods are further divided into three types: empirical / semi-empirical models, equivalent circuit models, and electrochemical mechanism models. Empirical / semi-empirical models typically construct relationships between battery capacity and aging indicators such as temperature and average state of charge (SOC) based on extensive experimental data. They generally use polynomial and exponential functions for fitting to establish the model. Their drawbacks are that they are generally only applicable to the same type of battery, and the models are simple with relatively large errors. Equivalent circuit models characterize the external characteristics of the battery's internal electrochemical reactions by combining equivalent circuit elements such as power supply, resistance, and capacitance. Common models include multi-order RC models and Rint models, but the accuracy of equivalent circuit models is limited, and their performance may be poor under extreme operating conditions or constant current conditions. Electrochemical mechanism models use multiple partial differential equations to calculate the battery's aging mechanism reaction process. These are derived from the battery's working principle and can describe the internal electrochemical kinetics of the battery. However, this method generally involves large computational loads and lacks simplified battery pack models, making it difficult to implement in battery packs. Data-driven methods rely on one-dimensional time-series data features of the battery, such as voltage, current, and temperature, as input into the model. However, these simple voltage and current features cannot fully represent the information about battery degradation. Furthermore, these methods are prone to problems such as overly deep networks, long training times, and overfitting.

[0057] In view of this, this application provides a battery health status assessment method. This battery health status assessment method can be applied to a terminal, a server, or software running on a terminal or server. The terminal can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited to these. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0058] Reference Figure 1 This invention provides a method for assessing battery health status, comprising:

[0059] S101. Perform energy feature extraction processing on the target battery to obtain sequential energy features;

[0060] S102. Perform data conversion processing on the sequence energy features to obtain multidimensional energy features;

[0061] S103. Input the multidimensional energy characteristics into the state assessment model to perform battery health state assessment processing to obtain the health state of the target battery. The state assessment model is a hybrid model based on the fusion of deep learning and machine learning.

[0062] This invention employs energy feature extraction to assess the health status of a target battery. The energy feature is the integral of the product of voltage and current over time, a comprehensive feature encompassing both current and voltage. Compared to other single features, it better captures information about battery degradation, identifies voltage and current changes during aging, and is easier to obtain in practice. It's important to note that the target battery in this invention is a battery in the constant current (CC) charging stage. Health status analysis is performed on the battery in the CC charging stage by extracting energy features from this stage. These energy features are time-series energy features, not the energy features for the entire stage. The energy value is calculated for each sampling point obtained through linear interpolation. This invention can also assess the health status of batteries in the constant voltage (CV) charging stage and the equal voltage discharge interval (EDVI). The sequence energy features are then processed through data conversion to obtain multidimensional energy features. In this invention, the multidimensional energy features can be two-dimensional, and the data conversion process can use a sliding window to transform the energy data into two-dimensional data similar to an image. Finally, the multidimensional energy features are input into the state assessment model to estimate the battery's health state. The state assessment model is a hybrid model based on the fusion of deep learning and machine learning. In this embodiment of the invention, the deep learning adopts the Residual Network model (ResNet) and the machine learning adopts the Support Vector Machine model (SVR). Combining deep learning and machine learning avoids the problem of overfitting due to too many parameters in the fully connected layers of deep learning, and also improves the training speed in terms of time.

[0063] As a further optional implementation, in step S101 above, the energy feature extraction process of the target battery to obtain sequential energy features includes:

[0064] Determine the sampling range of the target battery;

[0065] The target battery is sampled at equal intervals according to the sampling interval to obtain voltage data and current data.

[0066] The voltage and current data are linearly interpolated according to the energy characteristic calculation formula to obtain the sequence energy characteristics.

[0067] In this embodiment of the invention, the sampling interval of the target battery is determined by determining the starting voltage and ending voltage of the battery during constant current charging. In this embodiment, the initial charging cutoff voltage is set to 2V to 3.6V. However, it is difficult to guarantee that all battery cycles will reach these two voltages. To solve this problem, a voltage range is set so that all battery cycles can acquire voltages within this range; this voltage range is the sampling interval. Under this premise, the voltage range is maximized as much as possible to facilitate richer data in subsequent feature extraction. In this embodiment, the initial voltage and ending voltage are set to 2.8V and 3.494V, respectively. To address the problem of unequal time intervals between sampling points, this embodiment performs equal-interval sampling processing on the target battery according to the sampling interval, dividing the initial and ending voltage intervals into 100 equal points, and sampling voltage and current data corresponding to each sampling point. Finally, linear interpolation processing is performed on the voltage and current data according to the energy feature calculation formula to obtain the sequence energy features.

[0068] This invention uses energy as a health indicator for estimating battery health. Energy is the integral of voltage multiplied by current over time, comprehensively reflecting the energy changes in battery capacity during charging. (Refer to...) Figure 2 The voltage-time variation during the charging process is shown. It can be seen that as the battery ages, the voltage curve shifts to the left, meaning the voltage rises faster. In the voltage-time curve, as the number of battery cycles increases, the integral of the voltage-time curve also increases. Under constant current conditions, the integral of voltage over time equals energy, so energy can reflect the degree of battery aging to some extent. Using time as the horizontal axis and energy as the vertical axis makes it difficult to obtain richer information about battery cycles because the change in energy over time is essentially linear, which is not ideal for subsequent capacity estimation. Therefore, this embodiment of the invention uses voltage as the horizontal axis, as... Figure 3 As shown, energy is used as the vertical axis to explore richer information about battery aging.

[0069] As a further optional implementation, the step of performing linear interpolation processing on the voltage data and the current data according to the energy characteristic calculation formula to obtain the sequence energy characteristics includes:

[0070] Obtain the sampling time, sampling voltage, and sampling current corresponding to each sampling point of the linear interpolation;

[0071] The energy characteristics of the sequence are calculated by multiplying the sampling voltage and the sampling current according to the energy characteristic calculation formula and integrating them with the sampling time.

[0072] In this embodiment of the invention, by obtaining the sampling time, sampling voltage, and sampling current corresponding to each sampling point of the linear interpolation, and performing corresponding calculations according to the energy feature calculation formula, the energy value of each sampling point can be obtained. A sequence energy feature is then constructed from multiple energy values. The energy feature calculation formula is as follows:

[0073]

[0074] In the formula, I CC V represents the current used in constant current charging. CC The voltage represents constant current charging, E represents energy value, t represents sampling time, and t1 and t2 represent the start sampling point and end sampling point, respectively. In this embodiment of the invention, the sampling time is very short, so that the energy value is calculated once at each constant current sampling time point, thereby obtaining a series of energy values ​​and obtaining the sequence energy characteristics.

[0075] As a further optional implementation, the data transformation processing of the sequence energy features to obtain multidimensional energy features includes:

[0076] A one-dimensional energy characteristic curve is constructed based on the sequence energy characteristics. The one-dimensional energy characteristic curve is a curve with voltage as the horizontal axis and energy as the vertical axis.

[0077] The one-dimensional energy characteristic curve is sampled using a sliding window to obtain multi-dimensional energy characteristics.

[0078] In this embodiment of the invention, reference is made to Figure 3 A one-dimensional energy characteristic curve is constructed based on the sequence energy characteristics. This one-dimensional energy characteristic curve is a curve with voltage as the horizontal axis and energy as the vertical axis. (Refer to...) Figure 4 In this embodiment of the invention, a sliding window method is used to convert one-dimensional energy features into a multi-dimensional energy feature input format. In this embodiment, the multi-dimensional energy features use a two-dimensional energy feature format of 100*500*1 as the model input, where 100 refers to 100 windows and 500 refers to each window sampling 500 time-series energy data points. In a feasible embodiment, refer to... Figure 5 The window starts at an initial voltage of 2.8V, with a window length of 0.1V. Each shift involves a voltage increment of 0.006V, and after 99 shifts, 100 windows are obtained, eventually reaching an ending voltage of 3.494V. Finally, the energy characteristics of the entire one-dimensional charging curve are divided into 100 windows, and converted into a 100*500*1 two-dimensional feature by sampling 500 energy values ​​corresponding to each voltage. It's important to note that the 500 refers to sampling 500 energy values ​​corresponding to each voltage within each window, ensuring that each window yields 500 energy sequences.

[0079] As a further optional implementation, the step of inputting the multidimensional energy characteristics into the state assessment model for battery health state assessment to obtain the health state of the target battery includes:

[0080] Input the multidimensional energy characteristics into the state evaluation model;

[0081] The first module of the state assessment model performs feature extraction processing on the multidimensional energy features to obtain the semantic information of the multidimensional energy features;

[0082] The health status of the target battery is obtained by performing regression prediction processing on the semantic information through the second module of the state assessment model.

[0083] In this embodiment of the invention, the state assessment model adopts the ResNet-SVR model to estimate the battery health state. ResNet, as the first stage or first module of the model, is used for feature extraction, extracting high semantic information about energy features. The SVR layer replaces the fully connected layer and serves as the second module for regression prediction, combining the advantages of both deep learning and machine learning. Because the residual block contains shortcut connections, information spanning several layers is directly passed to the next layer through summation, thus retaining more information. Compared with other traditional neural networks, using ResNet in the first stage is more beneficial for feature extraction. Support Vector Regression is a machine learning algorithm used to analyze data regression problems. The input vector is mapped to a high-dimensional space, and a hyperplane is found in this high-dimensional space. For regression problems, the distance between the hyperplane and the farthest sample point must be minimized. This allows the hyperplane to fit the data, transforming a nonlinear problem into a linear problem in a high-dimensional space.

[0084] As a further optional implementation, before inputting the multidimensional energy features into the state assessment model for battery health state assessment, the method further includes pre-training the state assessment model, specifically including:

[0085] Build a state assessment model based on deep learning and machine learning;

[0086] By selecting multiple training batteries from a large battery dataset;

[0087] The training battery is subjected to multidimensional energy feature extraction processing to obtain a training dataset;

[0088] The training dataset is input into the state evaluation model, and the state evaluation model is trained by cross-validation.

[0089] In this embodiment of the invention, the MIT Battery Dataset is first introduced. MIT is a large dataset released by MIT, containing 124 lithium iron phosphate batteries, 72 different charging strategies, and cycle lives ranging from 150 to 2300. This embodiment selects seven of these batteries as training batteries. Their charging strategy is the same: '5.3C (54%) - 4C - newstructure'. This means that the battery is first charged at a constant current rate of 5.3C. When the state of charge (SOC) reaches 54%, it continues to be charged at a constant current rate of 4C. When the SOC reaches 80%, it is discharged at a constant current rate of 4C until approximately 80% of the battery capacity is recovered. Then, multidimensional energy feature extraction processing is performed on the training batteries to obtain a training dataset. The training dataset includes 100*500*1 energy features under multiple cycles. By using the 100*500*1 energy features under one cycle as input to the state assessment model, the output is the battery's state of health (SOH) under that cycle. Each cycle is one sample. The dataset has 7 batteries, for a total of 7273 cycles, i.e., 7273 samples. The training method uses cross-validation, i.e., one battery is used for testing, and the remaining batteries are used for training.

[0090] As a further optional implementation, the step of constructing a state evaluation model based on deep learning and machine learning includes:

[0091] Obtain a deep learning-based residual network model and a machine learning-based support vector machine model;

[0092] The fully connected layers of the residual network model are replaced according to the support vector machine model to obtain the state evaluation model.

[0093] In this embodiment of the invention, a ResNet-50 residual network model based on deep learning is used to extract high-dimensional semantic features of energy. The ResNet-50 residual network model is a 50-layer deep neural network with three fully connected layers at the end, used to combine learned features for prediction. This embodiment replaces the original three fully connected layers with a single SVR regression layer, which is more adaptable to small sample battery data. With small sample data, deep neural networks are more prone to overfitting. This embodiment uses ResNet to extract feature vectors with higher semantic information, reducing the dimensionality of the data and the number of model parameters, thereby alleviating overfitting. Furthermore, SVR is faster than fully connected layers. Since fully connected layers have a large number of parameters, converting them to SVR reduces computational complexity. Generally, deep neural networks are sensitive to noise; using an SVR model makes the model more resistant to noise, thus improving its robustness.

[0094] In one embodiment of the present invention, the starting and ending voltages of the battery are first determined during the charging process, and the energy value corresponding to each voltage sampling during the charging process is calculated. Then, 500 points are taken at equal intervals between these two voltages, and the energy value corresponding to each voltage is calculated by linear interpolation. At this point, an EV curve with voltage as the base and energy as the vertical axis is obtained, along with the energy values ​​corresponding to the 500 voltages, which constitute the time-series energy feature. Next, a sliding window is used during the input process to transform a cyclic one-dimensional time-series energy feature into two-dimensional image data, which is then fed into the ResNet-SVR network. After being input into the model, the battery health state estimation result can be obtained. Furthermore, the method was validated on the MIT dataset, showing good performance. (See reference...) Figures 6 to 9 This invention uses the MAE (Mean Absolute Error) metric to evaluate the effectiveness of the experiment. Training method: One battery is randomly selected for testing, and the remaining batteries are used for training. Therefore, this invention can obtain 7 results on the dataset. The minimum batch size is 15, the iteration is 50 epochs, and the learning rate is 0.001. This invention also conducted an SVR ablation experiment and demonstrated the effectiveness of SVR through experimental effect comparison. The best experimental result achieved an MAE of 0.0016, and the MAE of all batteries did not exceed 0.0102, with an average MAE of 0.0050. From the SVR ablation experiment, according to the MAE metric, the SVR regression prediction experiment shows higher accuracy and smaller error in SOH estimation. Therefore, it can be concluded that SVR can effectively replace fully connected layers.

[0095] This invention also provides a battery health status assessment system, comprising:

[0096] The first module is used to extract energy features from the target battery to obtain sequential energy features;

[0097] The second module is used to perform data conversion processing on the sequence energy features to obtain multidimensional energy features;

[0098] The third module is used to input the multidimensional energy characteristics into the state assessment model for battery health state assessment to obtain the health state of the target battery. The state assessment model is a hybrid model based on the fusion of deep learning and machine learning.

[0099] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0100] and Figure 1 Corresponding to the method described above, this embodiment of the invention also provides an electronic device, including a processor and a memory; the memory is used to store a program; the processor executes the program to implement the method described above.

[0101] and Figure 1 Corresponding to the method described above, embodiments of the present invention also provide a computer-readable storage medium storing a program that is executed by a processor to implement the method described above.

[0102] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform... Figure 1 The method shown.

[0103] In summary, the embodiments of the present invention have the following advantages:

[0104] 1. The present invention proposes to estimate the battery health status using energy characteristics. The energy of the sampled battery for each charging is calculated by voltage and current, and the energy is regarded as a function of voltage to obtain the sequence energy characteristics. This makes it easier to distinguish the differences between different cycles and thus see the trend of battery aging.

[0105] 2. Compared with general feature input, the embodiments of the present invention transform one-dimensional energy features into two-dimensional features similar to image data, fully explore the spatiotemporal changes of energy features, and are more conducive to training deep neural networks and more suitable for the input format of convolutional neural networks.

[0106] 3. The state assessment model in this embodiment combines ResNet and SVR, combining the advantages of deep learning and machine learning. ResNet, as a type of convolutional neural network, extracts deeper semantic features of energy data during battery aging through convolution operations. Combined with SVR in the final regression prediction, it can more accurately and quickly estimate the battery health status.

[0107] 4. The embodiments of the present invention have better predictive performance. It can be seen that in the final battery health state estimation results, the predicted value and the actual value curve are very close. The best experimental result reached MAE of 0.0016, and the MAE of all batteries did not exceed 0.0102.

[0108] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.

[0109] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0110] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0111] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0112] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0113] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0114] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0115] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0116] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for assessing battery health status, characterized in that, The method includes: Energy feature extraction is performed on the target battery to obtain sequential energy features; The sequence energy features are subjected to data transformation processing to obtain multidimensional energy features; The multidimensional energy features are input into the state assessment model to perform battery health state assessment processing, thereby obtaining the health state of the target battery. The state assessment model is a hybrid model based on the fusion of deep learning and machine learning. The data transformation processing of the sequence energy features to obtain multidimensional energy features includes: A one-dimensional energy characteristic curve is constructed based on the sequence energy characteristics. The one-dimensional energy characteristic curve is a curve with voltage as the horizontal axis and energy as the vertical axis. The one-dimensional energy feature curve is sampled using a sliding window to obtain a multi-dimensional energy feature; the multi-dimensional energy feature is in 100 format. 500 The two-dimensional energy features of 1 are used as the input of the model, 100 refers to 100 windows, and 500 refers to sampling 500 time-series energy data in each window; Before inputting the multidimensional energy features into the state assessment model for battery health state assessment, the method further includes pre-training the state assessment model, specifically including: Build a state assessment model based on deep learning and machine learning; By selecting multiple training batteries from a large battery dataset; The training battery is subjected to multidimensional energy feature extraction processing to obtain a training dataset; The training dataset is input into the state evaluation model, and the state evaluation model is cross-validated and trained. The construction of the state evaluation model based on deep learning and machine learning includes: Obtain a deep learning-based residual network model and a machine learning-based support vector machine model; The fully connected layers of the residual network model are replaced according to the support vector machine model to obtain the state evaluation model.

2. The method according to claim 1, characterized in that, The energy feature extraction process for the target battery to obtain sequential energy features includes: Determine the sampling range of the target battery; The target battery is sampled at equal intervals according to the sampling interval to obtain voltage data and current data. The voltage and current data are linearly interpolated according to the energy characteristic calculation formula to obtain the sequence energy characteristics.

3. The method according to claim 2, characterized in that, The step of performing linear interpolation on the voltage data and the current data according to the energy characteristic calculation formula to obtain the sequence energy characteristics includes: Obtain the sampling time, sampling voltage, and sampling current corresponding to each sampling point of the linear interpolation; The energy characteristics of the sequence are calculated by multiplying the sampling voltage and the sampling current according to the energy characteristic calculation formula and integrating them with the sampling time.

4. The method according to claim 1, characterized in that, The step of inputting the multidimensional energy characteristics into the state assessment model for battery health status assessment to obtain the health status of the target battery includes: Input the multidimensional energy characteristics into the state evaluation model; The first module of the state assessment model performs feature extraction processing on the multidimensional energy features to obtain the semantic information of the multidimensional energy features; The health status of the target battery is obtained by performing regression prediction processing on the semantic information through the second module of the state assessment model.

5. A battery health status assessment system, wherein the system is applied to the battery health status assessment method as described in claim 1, characterized in that, The system includes: The first module is used to extract energy features from the target battery to obtain sequential energy features; The second module is used to perform data conversion processing on the sequence energy features to obtain multidimensional energy features; The third module is used to input the multidimensional energy characteristics into the state assessment model for battery health state assessment to obtain the health state of the target battery. The state assessment model is a hybrid model based on the fusion of deep learning and machine learning.

6. An electronic device, characterized in that, The electronic device includes a memory and a processor; The memory is used to store programs; The processor executes the program to implement the method of any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 4.

Images