A battery full life cycle state of charge estimation method and system

Abstract

This invention discloses a method and system for estimating the state of charge (SOC) throughout the entire battery lifecycle. Addressing the problems of existing data-driven methods lacking understanding of battery physics and failing to dynamically update actual capacity, leading to accumulated SOC estimation errors over the entire lifecycle, this method constructs an SOC estimation feature set by fusing statistical features and equivalent circuit models to identify parameters online, and establishes an SOC estimation model based on a feedforward neural network. Simultaneously, based on the aging characteristics of the incremental capacity curve, an aging feature set is extracted to construct a capacity estimation neural network model, enabling online updates of the battery's actual capacity. By feeding the updated capacity back to the SOC estimation model in real time, its capacity benchmark is dynamically corrected, allowing the SOC estimation to adapt to the battery aging process, ultimately achieving high-precision and consistent SOC estimation throughout the entire battery lifecycle, from initial use to end of life. This invention improves the accuracy and robustness of battery state estimation, providing a reliable basis for battery management systems.

Description

Technical Field

[0001] This invention belongs to the technical field of lithium-ion battery state of charge estimation and battery capacity estimation, specifically a method and system for estimating the state of charge of a battery throughout its entire life cycle. Background Technology

[0002] Lithium-ion batteries, due to their high energy density and long cycle life, have become the core power source for electric vehicles. Their performance directly affects the vehicle's range, energy management strategy, and safety reliability. State of Charge (SOC), as a key parameter characterizing the remaining usable capacity of a battery, is one of the core estimation targets of battery management systems. However, lithium-ion batteries exhibit highly nonlinear electrochemical characteristics, making direct SOC measurement difficult. Furthermore, SOC is influenced by a combination of factors such as load current, operating temperature, and aging degree, making high-precision, full-lifecycle SOC estimation a challenge to date.

[0003] Currently, State of Charge (SOC) estimation methods can be broadly categorized into two types: model-based methods and data-driven methods. Model-based methods infer the internal state of the battery by establishing a mathematical model. Equivalent circuit models are widely used due to their simple structure and computational efficiency, but their model parameters often depend on specific operating conditions and struggle to reflect the internal electrochemical mechanisms of the battery. While electrochemical models can theoretically describe the internal processes of the battery, their complexity, numerous parameters, and high computational burden limit their engineering applicability. Data-driven methods, such as Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), and Gaussian Process Regression (GPR), do not rely on precise physical models but instead learn the mapping relationship between inputs and outputs through data, achieving high estimation accuracy on specific datasets. However, these methods are inherently dependent on statistical laws, lacking a clear expression of the internal physical mechanisms and exhibiting weak interpretability. More importantly, most existing methods typically assume a fixed initial battery capacity when estimating SOC, failing to adequately consider the continuous capacity decay of the battery throughout its lifespan. This leads to an accumulation of SOC estimation errors as the battery ages, making it difficult to achieve consistent and accurate estimations throughout the entire process from new battery to aging and disposal.

[0004] Therefore, how to combine the physical degradation mechanism of batteries with data-driven methods and achieve online adaptive updates of capacity parameters has become a key technical problem that needs to be solved to improve the accuracy and robustness of full life cycle SOC estimation. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a method and system for estimating the state of charge (SOC) of a battery throughout its entire life cycle. By incorporating statistical features and identification parameters based on an equivalent circuit model into the model input, a neural network-based SOC estimation model is constructed. A battery capacity estimation model based on IC curve aging characteristics is then introduced for capacity updates. The updated capacity is input into the SOC estimation model, ultimately achieving SOC estimation throughout the entire battery life cycle, providing theoretical guidance for the healthy operation of battery systems and electric vehicles.

[0006] In a first aspect, the present invention provides a method for estimating the state of charge (SOC) of a battery throughout its entire life cycle, comprising the following steps: S1. Constructing the SOC estimation feature set: Based on the original voltage and current time series data of battery operation, construct a SOC estimation feature set including three types of statistical features: sliding average, sliding standard deviation, and first-order difference, as well as the ohmic internal resistance, polarization internal resistance, polarization capacitance, and polarization voltage parameters obtained online based on the second-order equivalent circuit model. S2. Establish and run the SOC estimation model: Use the SOC estimation feature set constructed in step S1 as input, input it into the SOC estimation model constructed based on the feedforward neural network, and output the SOC estimation value at the current time. S3. Construct an aging feature set for capacity estimation: Based on the incremental capacity curve during the entire battery life cycle, extract the amplitude, voltage position, and voltage span between characteristic peaks on the curve to construct an aging feature set for capacity estimation. S4. Perform capacity update and full-cycle SOC estimation: Using the aging feature set constructed in S3 as input, input it into the battery capacity estimation model constructed based on the feedforward neural network, and output the battery capacity estimate; feed the capacity estimate into the SOC estimation model in S2 to dynamically update its capacity benchmark, thereby achieving accurate SOC estimation throughout the battery's entire life cycle.

[0007] Specifically, the formulas for calculating the moving average, moving standard deviation, and first difference are shown in equations (1)-(3), respectively: ; ; ; in, represents the characteristic value at time t; N represents the sliding window length.

[0008] Specifically, the online parameter identification based on the second-order equivalent circuit model adopts the extended Kalman filter algorithm, and the terminal voltage equation and polarization voltage differential equation of the second-order equivalent circuit model are shown in equations (4) and (5): ; ; In the formula, Open circuit voltage, This is the electrochemical polarization voltage. The concentration polarization voltage, For ohmic internal resistance, For electrochemical polarization internal resistance, For concentration polarization capacitors, Electrochemically polarized capacitor, This is the concentration-polarized capacitor.

[0009] Specifically, the online parameter identification process for the extended Kalman filter includes: set up Initial values ​​of the parameters to be identified and Initial values ​​for the state parameters; Time updates are performed, including updating the SOC via ampere-hour integration and updating the polarization voltage according to equation (6); ; Here, the current time is k, and the current parameter is updated based on the parameter k-1 from the previous time; this includes updating the SOC through ampere-hour integration to obtain the current time. The identification parameters are set to be the same as the previous time step, and will be corrected and updated in subsequent steps. The measurement update includes: predicting the current terminal voltage according to equation (7); ; In the formula, OCV is obtained through the current time. Correspondingly obtained; Calculate the Kalman gain matrix at the current time according to equation (8); ; In the formula, Let be the observation matrix, and represent the partial derivatives with respect to the state variables. To measure the noise variance; Based on the actual voltage at the current moment Calculate voltage error; ; And update and correct the identification parameters and state parameters according to equation (10); .

[0010] Specifically, in S2, the SOC estimation model is a feedforward neural network with an input layer, two hidden layers, and an output layer; The first hidden layer uses the ReLU activation function to activate nonlinear changes, as shown in equation (11): ; After the activation function, normalization is performed on each batch of data; Using Dropout's regularization method, a certain number of neurons are randomly set to zero in the current training step. A residual connection is introduced, a linear mapping branch is added to the second hidden layer, and it is added to the output of the second layer after ReLU transformation; The output layer is set to a single neuron, and the output is compressed to 0-1 using the Sigmoid activation function to output the SOC at the current time step.

[0011] Specifically, in S3, the incremental capacity curve is obtained according to the definition of equation (12); ; In the formula, For battery capacity, Battery voltage, and These represent minute changes in capacity and voltage, respectively.

[0012] The constructed aging feature set specifically includes: the amplitude of two peaks and their corresponding voltage values ​​on the IC curve, the amplitude of a trough and its corresponding voltage value, the voltage span between two peaks, and the voltage span between peaks and troughs.

[0013] Specifically, the first hidden layer of the battery capacity estimation model uses the ReLU activation function, followed by a Dropout layer; the second hidden layer uses the ReLU activation function; and the output layer uses a linear activation function to output the capacity estimate at the current time.

[0014] Secondly, a battery life-cycle state-of-charge estimation system is provided, the system comprising: The feature set construction module is used to perform step S1 and construct a SOC estimation feature set containing statistical features and online identification model parameters; The SOC estimation module is used to execute step S2, storing and running the feedforward neural network SOC estimation model. The aging feature extraction module is used to perform step S3, extract features from the incremental capacity curve, and construct an aging feature set for capacity estimation. The capacity estimation and feedback module is used to execute step S4, store and run the feedforward neural network capacity estimation model, and feed back the output capacity estimate to the SOC estimation module.

[0015] Thirdly, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0016] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0017] The beneficial effects of this invention are: This invention enables state-of-charge (SOC) estimation throughout the entire battery lifecycle. Based on the method described herein, firstly, statistical features are constructed from the original dataset. Then, online EKF parameter identification is performed based on a second-order equivalent circuit model. The obtained statistical features and identified parameters are used as model inputs, and a SOC estimation model is built based on a neural network. Next, a battery capacity estimation model based on IC curve aging features is introduced for capacity updates. The updated capacity is then input into the SOC estimation model, ultimately achieving SOC estimation throughout the entire battery lifecycle. This provides an important theoretical basis for the long-term effective operation of battery systems and electric vehicles. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the method described in this invention; Figure 2 This is a circuit structure diagram of the second-order equivalent circuit model of the present invention; Figure 3 This is a structural logic diagram of the SOC estimation model based on neural networks in this invention; Figure 4 This is a graph showing the change of IC curve with aging in an embodiment of the present invention; Figure 5 This is a structural logic diagram of the battery capacity estimation model based on neural networks of the present invention; Figure 6 The following is a logical diagram of the full lifecycle SOC estimation model for capacity updates added to this invention; Figure 7 This is a diagram showing the SOC estimation results of an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, advantages and features of the present invention more apparent, the present invention will be further described in detail below with reference to specific embodiments.

[0020] like Figure 1 The diagram shows a flowchart of a battery life-cycle state-of-charge estimation method according to the present invention, which specifically includes the following steps: Step 1: Construct a SOC estimation feature set based on statistical features and equivalent circuit model identification parameters.

[0021] For the original voltage, current, and capacity time-series dataset, a sliding window length of N is set, and data from the previous N cycles are extracted to establish three types of statistical features. These include a moving average, which averages the voltage features within the current window to capture short-term voltage trend changes within the sliding window, as expressed in equation (1): ; In the formula, express t The characteristic value at time step.

[0022] The sliding standard deviation measures the fluctuation of data within a sliding window, identifying the degree of load voltage change under short-term trends, as expressed in equation (2): ; The first-order difference reveals the dynamic nature of the timing characteristics, characterizing the speed and direction of voltage and current changes, as expressed in equation (3): ; Then, using the same original dataset, online EKF parameter identification based on a second-order equivalent circuit model is performed. The circuit structure diagram of the second-order equivalent circuit model is shown below. Figure 2 As shown, the terminal voltage equation and polarization voltage differential equation of this model are shown in equations (4) and (5): ; ; In the formula, Open circuit voltage, This is the electrochemical polarization voltage. The concentration polarization voltage, For ohmic internal resistance, For electrochemical polarization internal resistance, For concentration polarization capacitors, Electrochemically polarized capacitor, This is the concentration-polarized capacitor.

[0023] After the model is built, the known voltage, current, temperature, and OCV are input, and the EKF online parameter identification is performed according to the model formula. The process includes: set up Initial values ​​of the parameters to be identified and Initial values ​​for the state parameters.

[0024] Step 1: Let the current time be... Predicting current parameters based on parameters from the previous time step, i.e., time updating. This includes updating the current SOC through ampere-hour integration. The polarization voltage is updated as shown in equation (6). The identification parameters are temporarily set to be the same as the previous time step, and will be corrected and updated later.

[0025] ; Step 2: Predict the current terminal voltage based on the known OCV and current, as shown in Equation (7); and calculate the Kalman gain matrix at the current moment. As shown in equation (8).

[0026] ; In the formula, OCV is obtained through the current time. The corresponding result is obtained.

[0027] ; In the formula, Let be the observation matrix, and represent the partial derivatives with respect to the state variables. To measure the noise variance.

[0028] Step 3: Update and correct the parameters at the current moment. First, based on the actual voltage at the current moment... The voltage error is calculated as shown in Equation (9); then the identification parameters and state parameters are updated as shown in Equation (10).

[0029] ; ; Finally, the statistical features and the identified parameters are used to construct a feature set as input to the SOC estimation model.

[0030] In this embodiment, the sliding window length N is 5, and the statistical features selected are the voltage sliding average, voltage sliding standard deviation, and the first-order difference between voltage and current. In EKF online parameter identification, voltage, current, and temperature are known. The OCV-SOC correspondence is obtained through HPPC testing, and the initial values ​​of the parameters to be identified are set. The state parameters are Select Polarization voltage is incorporated into the feature set construction.

[0031] Step 2: Based on the constructed dataset input, build a neural network model for SOC estimation.

[0032] The model input is the feature set constructed in step one. A feedforward neural network is used for modeling, employing a three-layer structure: an input layer, two hidden layers, and an output layer. The logical structure diagram is as follows: Figure 3 As shown. The first hidden layer introduces the ReLU function to activate nonlinear changes. Its expression is shown in equation (11): ; After the activation function, the model introduces batch normalization to normalize each batch of data. Then, Dropout regularization is applied to randomly set the output of 20% of the neurons in the current training step to zero. The second hidden layer still uses the ReLU function and introduces a residual connection, that is, a linear mapping branch is added to the second hidden layer and added to the output of the second layer after ReLU transformation.

[0033] The final output layer is set to a single neuron, and the Sigmoid activation function is used to compress the output to 0-1, outputting the SOC at the current time step.

[0034] In this embodiment, grid search is used to find the optimal hyperparameters, and the search space includes: the model learning rate. The batch size of neurons input each time and number of training rounds The loss function is the mean squared error (MSE), and the EarlyStopping mechanism is introduced to avoid overfitting.

[0035] Step 3: Based on the changes in IC curves with aging, construct an aging feature set to characterize battery aging.

[0036] The IC curve is an important curve obtained by differentiating the relationship between voltage and capacity during battery charging and discharging. Its definition formula is (12): ; In the formula, For battery capacity, Battery voltage, and These represent minute changes in capacity and voltage, respectively.

[0037] As batteries are used and age, the overall shape and characteristic peaks of the IC curve will change. Specifically, with capacity decay, the height of the peaks and troughs decreases, their positions shift, and the voltage span increases.

[0038] Based on the changing characteristics of the IC curve, the height of the peaks and troughs reflects the degree of degradation of the active materials in the battery; the voltage positions of the peaks and troughs reflect the increase and irreversible change of the battery's internal resistance; and the voltage span between peaks and troughs reflects the battery's reaction consistency and polarization effect. Finally, based on the characterized aging characteristics, the heights and voltage positions of the two peaks, the heights and voltage positions of the troughs, and the voltage spans between the two peaks and between the peaks and troughs of the IC curve are extracted to construct an aging feature set.

[0039] In this embodiment, according to as follows Figure 4The graph shown shows the IC curve changing with aging. In each cycle, nine aging feature sets derived from the IC curve are extracted.

[0040] Step 4: Based on the established aging feature set as input, construct a battery capacity estimation model to update the capacity, input the updated capacity into the SOC estimation model, and finally realize the SOC estimation of the entire life cycle.

[0041] The pre-constructed aging feature set is input, and a battery capacity estimation model is still built based on a feedforward neural network. The structure logic diagram is as follows: Figure 5 As shown. The first hidden layer still uses the ReLU activation function, followed by a Dropout layer; the second hidden layer also uses the ReLU activation function; the output layer uses a linear activation function to output the capacity estimate at the current time step.

[0042] After obtaining the capacity estimate, the updated capacity is fed back to the SOC estimation model in real time, dynamically correcting the available capacity in the current cycle. This ensures that the SOC estimation results always adapt to the capacity decay characteristics caused by battery aging, thereby achieving accurate SOC estimation throughout the entire battery lifecycle from its introduction to the end of its lifespan. The logic diagram of the full-lifecycle SOC estimation model with capacity updates is shown below. Figure 6 As shown.

[0043] In this embodiment, the battery capacity estimation model uses RMSPE, MAXPE, and MAPE to evaluate the battery capacity estimation performance. The error results for different test sets are shown in Table (1). The RMSPE and MAPE of the three experiments are all less than 1%, indicating that the overall estimation error of the model is small; while MAXPE is also kept below 3%, indicating that the maximum error is within a reasonable range. After feeding the updated capacity back to the SOC estimation model, the curves of some true values ​​and estimated values ​​of SOC are shown below. Figure 7 As shown, it is proven that SOC estimation for the entire life cycle is achieved, and the results are good on different temperature datasets, demonstrating broad applicability.

[0044] Table 1. Error results under different test sets Test dataset RMSPE (%) MAXPE (%) MAPE (%) Battery_38 0.835 2.923 0.681 Battery_37 0.919 2.968 0.783 Battery_36 0.729 2.695 0.561 ;

[0045] Based on the same inventive concept, the present invention also provides a battery life-cycle state of charge estimation system, the system comprising: The feature set construction module is used to perform step S1 and construct a SOC estimation feature set containing statistical features and online identification model parameters; The SOC estimation module is used to execute step S2, storing and running the feedforward neural network SOC estimation model. The aging feature extraction module is used to perform step S3, extract features from the incremental capacity curve, and construct an aging feature set for capacity estimation. The capacity estimation and feedback module is used to execute step S4, store and run the feedforward neural network capacity estimation model, and feed back the output capacity estimate to the SOC estimation module.

[0046] It should be noted that any process or method description in the embodiments can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as should be understood by those skilled in the art to which the embodiments of the invention pertain.

[0047] It should be noted that the logic and / or steps in the embodiments, 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. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), 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). Alternatively, the computer-readable medium may 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 a computer memory.

[0048] 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, 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.

[0049] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0050] Furthermore, in the embodiments of the present invention, the functional modules can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The aforementioned storage medium can be a read-only memory, a hard disk, or an optical disk, etc.

[0051] The above embodiments have provided a detailed description of the technical solutions of the present invention. Obviously, the present invention is not limited to the described embodiments. Based on the embodiments of the present invention, those skilled in the art can make various changes, but any changes that are equivalent or similar to the present invention fall within the scope of protection of the present invention. Contents not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for estimating the state of charge (SOC) throughout the entire lifespan of a battery, characterized in that, Includes the following steps: S1. Constructing the SOC estimation feature set: Based on the original voltage and current time series data of battery operation, construct a SOC estimation feature set including three types of statistical features: sliding average, sliding standard deviation, and first-order difference, as well as the ohmic internal resistance, polarization internal resistance, polarization capacitance, and polarization voltage parameters obtained online based on the second-order equivalent circuit model. S2. Establish and run the SOC estimation model: Use the SOC estimation feature set constructed in step S1 as input, input it into the SOC estimation model constructed based on the feedforward neural network, and output the SOC estimation value at the current time. S3. Construct an aging feature set for capacity estimation: Based on the incremental capacity curve during the entire battery life cycle, extract the amplitude, voltage position, and voltage span between characteristic peaks on the curve to construct an aging feature set for capacity estimation. S4. Perform capacity update and full-cycle SOC estimation: Using the aging feature set constructed in S3 as input, input it into the battery capacity estimation model constructed based on the feedforward neural network, and output the battery capacity estimate. The capacity estimate is fed back into the SOC estimation model of S2 to dynamically update its capacity benchmark, thereby achieving accurate SOC estimation over the entire battery life cycle.

2. The method according to claim 1, characterized in that, The formulas for calculating the moving average, moving standard deviation, and first difference are as follows: ； ； ； in, represents the characteristic value at time t; N represents the sliding window length.

3. The method according to claim 1, characterized in that, The online parameter identification based on the second-order equivalent circuit model adopts the extended Kalman filter algorithm. The terminal voltage equation and polarization voltage differential equation of the second-order equivalent circuit model are shown in equations (4) and (5): ； ； In the formula, Open circuit voltage, This is the electrochemical polarization voltage. The concentration polarization voltage, For ohmic internal resistance, For electrochemical polarization internal resistance, For concentration polarization capacitors, Electrochemically polarized capacitor, This is the concentration-polarized capacitor.

4. The method according to claim 3, characterized in that, The execution flow of the extended Kalman filter in the online parameter identification process includes: set up Initial values ​​of the parameters to be identified and Initial values ​​for the state parameters; Time updates are performed, including updating the SOC via ampere-hour integration and updating the polarization voltage according to equation (6); ； Here, the current time is k, and the current parameter is updated based on the parameter k-1 from the previous time; this includes updating the SOC through ampere-hour integration to obtain the current time. The identification parameters are set to be the same as the previous time step, and will be corrected and updated in subsequent steps. The measurement update includes: predicting the current terminal voltage according to equation (7); ； In the formula, OCV is obtained through the current time. Correspondingly obtained; Calculate the Kalman gain matrix at the current time according to equation (8); ； In the formula, Let be the observation matrix, and represent the partial derivatives with respect to the state variables. To measure the noise variance; Based on the actual voltage at the current moment Calculate voltage error; ； And update and correct the identification parameters and state parameters according to equation (10); 。 5. The method according to claim 1, characterized in that, In S2, the SOC estimation model is a feedforward neural network with an input layer, two hidden layers, and an output layer. The process is as follows: The first hidden layer uses ReLU activation for nonlinear changes, as shown in equation (11): ； After the activation function, normalization is performed on each batch of data; Using Dropout's regularization method, a certain number of neurons are randomly set to zero in the current training step. A residual connection is introduced, a linear mapping branch is added to the second hidden layer, and it is added to the output of the second layer after ReLU transformation; The output layer is set to a single neuron, and the output is compressed to 0-1 using the Sigmoid activation function to output the SOC at the current time step.

6. The method according to claim 1, characterized in that, In S3, the incremental capacity curve is obtained according to the definition of equation (12); ； In the formula, For battery capacity, Battery voltage, and These represent minute changes in capacity and voltage, respectively. The constructed aging feature set specifically includes: the amplitude of two peaks and their corresponding voltage values ​​on the IC curve, the amplitude of a trough and its corresponding voltage value, the voltage span between two peaks, and the voltage span between peaks and troughs.

7. The method according to claim 1, characterized in that, In S4, the first hidden layer of the battery capacity estimation model uses the ReLU activation function, followed by a Dropout layer; the second hidden layer uses the ReLU activation function; and the output layer uses a linear activation function to output the capacity estimate at the current time.

8. A battery life-cycle state-of-charge estimation system, characterized in that, The system for implementing the method as described in any one of claims 1 to 7 comprises: The feature set construction module is used to perform step S1 and construct a SOC estimation feature set containing statistical features and online identification model parameters; The SOC estimation module is used to execute step S2, store and run the feedforward neural network SOC estimation model; The aging feature extraction module is used to perform step S3, extract features from the incremental capacity curve, and construct an aging feature set for capacity estimation. The capacity estimation and feedback module is used to execute step S4, store and run the feedforward neural network capacity estimation model, and feed back the output capacity estimate to the SOC estimation module.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored on the memory, wherein the processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program thereon, which, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 7.

Images