Lithium ion battery charge state monitoring method, device and equipment and storage medium
By extracting the characteristic frequency impedance parameters of lithium-ion batteries and performing dimensionality reduction, a state-of-charge estimation model is constructed, which solves the problem of low accuracy in monitoring the state of charge of lithium-ion batteries under dynamic operating conditions and realizes high-precision online real-time monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEBEI UNIV OF TECH
- Filing Date
- 2026-03-10
- Publication Date
- 2026-05-12
Smart Images
Figure CN122017616A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of lithium-ion battery state of charge management technology, and in particular to a method, apparatus, device and storage medium for monitoring the state of charge of lithium-ion batteries. Background Technology
[0002] Currently, methods for estimating the state of charge (SOC) of lithium-ion batteries present diverse technical approaches. Among them, the open-circuit voltage method, a classic approach, is based on the principle that there is a definite functional relationship between the open-circuit voltage and the state of charge of the battery in its resting state; by measuring the open-circuit voltage under steady-state conditions, the battery's state of charge can be determined. In addition, the ampere-hour integration method calculates the change in charge by integrating the charging and discharging current over time, while data-driven methods rely on mining large amounts of historical operating data to build estimation models. Model-based methods simulate and extrapolate the internal state of the battery by establishing electrochemical models or equivalent circuit models.
[0003] However, the aforementioned existing technologies all have certain limitations in practical applications. Specifically, the open-circuit voltage method requires the battery to be in a static state for a long period of time to achieve voltage stability. This prerequisite is difficult to meet during actual battery operation, making it unsuitable for online monitoring under dynamic operating conditions. More importantly, lithium-ion batteries have a relatively flat voltage plateau in the middle region of their state of charge (SOC). Within this range, the voltage changes very little with the SOC, making it difficult for the open-circuit voltage method to accurately distinguish the SOC, thus reducing the estimation accuracy. Therefore, existing technologies cannot achieve high-precision online monitoring of the SOC of lithium-ion batteries under dynamic operating conditions. Summary of the Invention
[0004] This application provides a method, apparatus, device, and storage medium for monitoring the state of charge (SOC) of lithium-ion batteries. It can achieve high-precision real-time monitoring of the SOC of lithium-ion batteries based on online impedance spectroscopy. By optimizing characteristic frequencies and globally optimizing model parameters, it overcomes the problems of traditional open-circuit voltage methods being unable to monitor online and low accuracy in the voltage plateau region.
[0005] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, this application provides a method for monitoring the state of charge of a lithium-ion battery, the method comprising: To obtain electrochemical impedance spectroscopy data of lithium-ion batteries under different temperatures, aging conditions, and charging states; Based on the electrochemical impedance spectroscopy data, impedance parameters at characteristic frequencies related to the state of charge are extracted; The extracted impedance parameters are dimensionality reduced to obtain the dimensionality-reduced feature data. Based on the reduced-dimensional feature data, a state of charge estimation model is constructed; Acquire the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation; Based on the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model, the target state of charge of the lithium-ion battery is obtained.
[0006] In some possible implementations, the extraction of impedance parameters at characteristic frequencies related to the state of charge based on the electrochemical impedance spectroscopy data includes: Based on the electrochemical impedance spectroscopy data, the correlation coefficient between impedance parameters and state of charge at different frequencies is calculated; according to the magnitude of the correlation coefficient, a first characteristic frequency with high correlation to the state of charge is selected; based on the electrochemical impedance spectroscopy data and the first characteristic frequency, the impedance parameters at the first characteristic frequency are extracted.
[0007] In some possible implementations, the dimensionality reduction processing of the impedance parameters to obtain dimensionality-reduced feature data includes: Principal component analysis is performed on the impedance parameters to reduce their dimensionality, and the first principal component features are obtained; the first principal component features are used as the feature data after dimensionality reduction.
[0008] In some possible implementations, constructing a state of charge estimation model based on the dimensionality-reduced feature data includes: Using the dimensionality-reduced feature data, battery temperature data, and battery aging state data as inputs, and the state of charge of the lithium-ion battery as the output, an initial estimation model is constructed. An optimization algorithm is used to globally optimize the hyperparameters of the initial estimation model to obtain a first hyperparameter combination. Based on the first hyperparameter combination, a state of charge estimation model is trained.
[0009] In some possible implementations, acquiring the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation includes: Under static or dynamic operating conditions, a multi-frequency AC excitation signal is applied to the lithium-ion battery; response data of the lithium-ion battery to the multi-frequency AC excitation signal is collected; and based on the response data, the first multi-frequency electrochemical impedance spectroscopy data is calculated.
[0010] In some possible implementations, obtaining the target state of charge of the lithium-ion battery based on the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model includes: From the first multi-frequency electrochemical impedance spectroscopy data, a first impedance parameter corresponding to the first characteristic frequency is extracted; the extracted first impedance parameter is subjected to the dimensionality reduction processing to obtain first dimensionality-reduced feature data; the first dimensionality-reduced feature data is input into the state of charge estimation model to output the target state of charge of the lithium-ion battery.
[0011] In some possible implementations, the method further includes, before acquiring electrochemical impedance spectroscopy data of lithium-ion batteries at different temperatures, aging states, and states of charge: Perform standard charge-discharge cycles on the lithium-ion battery to determine its actual discharge capacity.
[0012] Secondly, this application provides a monitoring device for the state of charge of a lithium-ion battery, the device comprising: The acquisition module is used to acquire electrochemical impedance spectroscopy data of lithium-ion batteries at different temperatures, different aging states, and different states of charge; based on the electrochemical impedance spectroscopy data, impedance parameters at characteristic frequencies related to the state of charge are extracted; and the impedance parameters are subjected to dimensionality reduction processing to obtain dimensionality-reduced feature data. The calculation module is used to construct a state of charge estimation model based on the dimensionality-reduced feature data; acquire the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation; and obtain the target state of charge of the lithium-ion battery according to the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model.
[0013] Thirdly, this application provides a computing device, including a memory and a processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of the first aspects.
[0014] Fourthly, this application provides a computer-readable storage medium for storing a computer program for performing the method as described in any one of the first aspects.
[0015] Fifthly, this application provides a computer program product comprising one or more computer instructions, wherein when the computer instructions are executed by a computer, the computer performs the method as described in any one of the first aspects.
[0016] As can be seen from the above technical solution, this application has at least the following beneficial effects: In this application, electrochemical impedance spectroscopy (EIS) data of lithium-ion batteries under different temperatures, aging conditions, and states of charge are obtained; based on the EIS data, impedance parameters at characteristic frequencies related to the state of charge are extracted; the extracted impedance parameters are dimensionality-reduced to obtain dimensionality-reduced feature data; based on the dimensionality-reduced feature data, a state of charge estimation model is constructed; first multi-frequency EIS data of the lithium-ion battery during operation are obtained; and the target state of charge of the lithium-ion battery is obtained based on the first multi-frequency EIS data and the state of charge estimation model.
[0017] In existing technologies, the open-circuit voltage method requires the battery to be stationary for a long time and suffers from low estimation accuracy due to voltage plateau effects, making it impossible to achieve online monitoring under dynamic operating conditions. Therefore, this application reduces data redundancy and multicollinearity by extracting impedance parameters at characteristic frequencies related to the state of charge (SOC) and performing dimensionality reduction, thus lowering the model input dimension. By constructing a SOC estimation model and embedding it into the monitoring device, it achieves online real-time monitoring of the SOC of lithium-ion batteries under both static and dynamic operating conditions. Furthermore, by employing an optimization algorithm to globally optimize the model's hyperparameters, it improves the accuracy of SOC estimation and the model's generalization ability. Finally, by replacing traditional frequency sweep testing with multi-frequency electrochemical impedance spectroscopy, it shortens the measurement time and meets the real-time requirements of online monitoring. This application overcomes the problems of the traditional open-circuit voltage method's inability to perform online monitoring and its low accuracy in the voltage plateau region, achieving high-precision online real-time monitoring of the SOC of lithium-ion batteries.
[0018] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description
[0019] Figure 1 A flowchart illustrating a method for monitoring the state of charge of a lithium-ion battery, provided as an embodiment of this application; Figure 2 A schematic diagram of an impedance-state-of-charge monitoring device provided in an embodiment of this application; Figure 3An interpretation diagram of the impedance imaginary part PCA variance provided in this application embodiment; Figure 4 A flowchart illustrating a Gaussian process regression hyperparameter optimization method using an exponential triangular optimization algorithm, provided in this application embodiment; Figure 5 A diagram showing the SOC estimation results provided in the embodiments of this application; Figure 6 A schematic diagram of a lithium-ion battery state-of-charge monitoring device provided in an embodiment of this application; Figure 7 This is a schematic diagram of a computing device provided in an embodiment of this application. Detailed Implementation
[0020] The terms "first," "second," and "third," etc., used in this application specification and accompanying drawings are used to distinguish different objects, not to limit a specific order.
[0021] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.
[0022] To ensure clarity and conciseness in the description of the following embodiments, a brief introduction to the related technologies is given first: Currently, methods for estimating the state of charge (SOC) of lithium-ion batteries mainly include the open-circuit voltage method, the ampere-hour integration method, the data-driven method, the model-based method, and related methods based on electrochemical impedance spectroscopy. Among these, the open-circuit voltage method estimates the SOC using the functional relationship between the open-circuit voltage and the battery's SOC under quiescent conditions; the ampere-hour integration method calculates the change in charge by integrating the charging and discharging current over time; the data-driven method relies on mining large amounts of historical operating data to construct an estimation model; the model-based method infers the battery's internal state by establishing an electrochemical model or equivalent circuit model; and the electrochemical impedance spectroscopy-based method reflects the battery's internal electrochemical characteristics by measuring the AC impedance at different frequencies.
[0023] Existing technologies all have limitations in practical applications. Specifically, the open-circuit voltage method requires the battery to be in a static state for a long time to obtain a stable voltage, making it unsuitable for online monitoring under dynamic operating conditions. Furthermore, lithium-ion batteries have a flat voltage plateau in the middle region of their state of charge (SOC), within which the voltage changes very little with SOC, leading to a significant decrease in estimation accuracy. The ampere-hour integration method is sensitive to the initial SOC value, and the errors generated during integration accumulate over time, making effective correction impossible. Data-driven methods suffer from difficulties in feature selection and insufficient model generalization ability, and require massive amounts of data for offline training. Model-based methods involve numerous electrochemical model parameters and complex calculations, while equivalent circuit models face difficulties in parameter identification. Traditional methods based on electrochemical impedance spectroscopy use frequency sweeping to acquire impedance data, resulting in long detection times, and the battery state may change during measurement, failing to meet the needs of online real-time monitoring.
[0024] In view of this, embodiments of this application provide a method for monitoring the state of charge (SOC) of a lithium-ion battery. This method involves acquiring electrochemical impedance spectroscopy (EIS) data of the lithium-ion battery at different temperatures, aging states, and SOC states; extracting impedance parameters at characteristic frequencies related to the SOC and performing dimensionality reduction; constructing an SOC estimation model based on the dimensionality-reduced feature data; and then using real-time multi-frequency EIS data and the SOC estimation model to obtain the target SOC of the lithium-ion battery. This method is applied to electric vehicles and executed by the battery management system (BMS). The BMS, as an on-board processing device, uses real-time multi-frequency EIS data collected during vehicle operation, combined with the pre-constructed SOC estimation model, to estimate the SOC of the power battery online. The estimation results are then sent to the vehicle controller and instrument panel for vehicle energy management and range display.
[0025] To make the technical solution of this application clearer and easier to understand, the following describes a method for monitoring the state of charge of a lithium-ion battery according to an embodiment of this application, in conjunction with the accompanying drawings. Figure 1 As shown in the figure, this is a flowchart of a method for monitoring the state of charge of a lithium-ion battery according to an embodiment of this application.
[0026] The method includes: S201. Obtain electrochemical impedance spectroscopy data of lithium-ion batteries under different temperatures, different aging states, and different charging states.
[0027] Electrochemical impedance spectroscopy (EIS) is a testing method that characterizes the internal electrochemical properties of a battery by applying a small AC excitation signal to it and measuring the battery's impedance response at different frequencies. EIS data includes parameters across multiple dimensions, such as the real part of impedance, the imaginary part of impedance, the impedance magnitude, and the impedance phase angle.
[0028] State of Health (SOH) is used to characterize the degree of capacity decay of a lithium-ion battery relative to its initial capacity, reflecting the degree of battery aging.
[0029] State of charge (SOC) is used to characterize the percentage of a lithium-ion battery's current remaining usable capacity relative to its current total usable capacity, reflecting the battery's remaining charge.
[0030] Before performing step S201, the lithium-ion battery is subjected to a standard charge-discharge cycle to determine its actual discharge capacity.
[0031] A standard charge-discharge cycle refers to a complete charge and discharge cycle of a lithium-ion battery under specified environmental conditions and charge-discharge regimes. Standard charge-discharge cycles are used to activate the battery, stabilize its performance, and determine its actual usable capacity under the current conditions.
[0032] Actual discharge capacity refers to the total amount of electricity released by a lithium-ion battery when discharged from a fully charged state to the cutoff voltage under standard charge and discharge conditions, usually measured in ampere-hours (Ah) or milliampere-hours (mAh). Accurate measurement of actual discharge capacity is fundamental for subsequently determining the battery's aging state and state of charge.
[0033] Specifically, the lithium-ion battery is connected to the charge-discharge testing device to ensure reliable electrical connection. A temperature sensor is attached to the battery surface to monitor battery temperature. The battery is placed in a constant temperature environment to allow the battery temperature to reach the preset test temperature and remain stable. In this embodiment, a lithium iron phosphate battery is selected and subjected to 5 standard charge-discharge cycles at room temperature (25°C): charged at a constant current of 0.5C to 3.65V, then switched to constant voltage charging until the current drops to 0.05C; after resting, discharged at a constant current of 0.5C to 2.5V. During the discharge process, the discharge current and discharge time are recorded in real time.
[0034] The lithium-ion battery is charged with a constant current. When the battery voltage reaches the charging cutoff voltage, it switches to a constant voltage charging mode and continues charging until the charging current drops below the set threshold. At this point, the battery is considered to be fully charged and the state of charge reaches 100%.
[0035] After the battery has been left to stand for a period of time, the lithium-ion battery is discharged at a constant current until the battery voltage drops to the discharge cutoff voltage. During the discharge process, the discharge current and discharge time are recorded in real time.
[0036] Based on the discharge current and time data recorded during the discharge process, the total amount of electricity released by the battery during this discharge is calculated using the ampere-hour integration method, which represents the actual discharge capacity of the lithium-ion battery in its current state. The formula for calculating the actual discharge capacity is as follows:
[0037] In the formula, Let I(t) be the actual discharge capacity, I(t) be the discharge current at time t, and t be the total discharge time. In the case of discrete sampling, numerical integration can be used for calculation.
[0038] Repeat the above steps multiple times, typically 3-5 times, until the deviation between two consecutive measured actual discharge capacities is less than a preset threshold, indicating that the battery performance has stabilized. Take the average of the multiple measurements as the final actual discharge capacity of the lithium-ion battery.
[0039] Then proceed with step S201, the specific process of which is as follows: Electrochemical impedance spectroscopy (EIS) tests were performed on lithium-ion batteries under different temperature conditions, aging states, and states of charge using an impedance-state-of-charge (ESC) monitoring device. During the tests, multi-frequency AC excitation signals were applied to the lithium-ion batteries. In a preferred embodiment, the excitation current amplitude was set to 100 mA, with a frequency range covering 0.01 Hz to 10 kHz to ensure comprehensive excitation of the battery's impedance response in different electrochemical processes. The response data of the lithium-ion batteries to the multi-frequency AC excitation signals were collected, and electrochemical impedance spectroscopy data were calculated based on this data. The electrochemical impedance spectroscopy data includes the real part of impedance, the imaginary part of impedance, the impedance magnitude, and the impedance phase angle.
[0040] The impedance-state-of-charge (APC) monitoring device is a specialized device used to apply multi-frequency AC excitation signals to a lithium-ion battery, collect the response data of the lithium-ion battery to the multi-frequency AC excitation signals, calculate electrochemical impedance spectroscopy (EIS) data based on the response data, and process the EIS data to estimate the state of charge (SOC) of the lithium-ion battery. The APC includes an AC excitation signal generation unit, a response signal acquisition unit, an impedance calculation unit, and a SOC estimation unit. The APC is communicatively connected to a host computer, receives control commands from the host computer, and uploads the processed SOC data to the host computer for real-time display. This application provides a schematic diagram of an APC, as shown below. Figure 2As shown in the figure, the system includes a host computer 1, an impedance-state-of-charge monitoring device 2, a charge-discharge testing device 3, a constant temperature chamber 4, a temperature sensor 5, and a lithium-ion battery 6. The host computer 1 is connected to the impedance-state-of-charge monitoring device 2 and the charge-discharge testing device 3 to achieve control and data transmission. The impedance-state-of-charge monitoring device 2 is connected to the lithium-ion battery 6 to provide AC excitation signals and collect impedance data. The charge-discharge testing device 3 is connected to the lithium-ion battery 6 to achieve charge and discharge under different operating conditions. The constant temperature chamber 4 is used to provide a stable ambient temperature. The temperature sensor 5 is attached to the battery surface to monitor the battery temperature in real time.
[0041] By acquiring electrochemical impedance spectroscopy data of lithium-ion batteries at different temperatures (e.g., -10℃ to 45℃), different aging states (e.g., SOH decaying from 100% to 80%), and different states of charge (0% to 100%), an impedance characteristic database covering the entire battery life cycle and all operating conditions was established. This provides a comprehensive and reliable data foundation for subsequent characteristic frequency extraction and state of charge estimation model construction, ensuring the adaptability and generalization ability of the model under different usage conditions.
[0042] S202. Based on electrochemical impedance spectroscopy data, extract impedance parameters at characteristic frequencies related to the state of charge.
[0043] Characteristic frequencies are specific frequency points in the electrochemical impedance spectroscopy that are highly correlated with the state of charge (SOC) of a lithium-ion battery. At these frequency points, the impedance parameters are most sensitive to changes in the SOC.
[0044] Spearman's Rank Correlation Coefficient is a non-parametric correlation measure used to assess the strength of the monotonic correlation between two variables, without requiring the data to satisfy the normal distribution assumption.
[0045] Based on electrochemical impedance spectroscopy (EIS) data, the correlation coefficients between impedance parameters and state of charge at different frequencies are calculated. According to the magnitude of the correlation coefficients, a first characteristic frequency with a high correlation to the state of charge is selected. Based on the EIS data and the first characteristic frequency, the impedance parameters at the first characteristic frequency are extracted. The specific process is as follows: Based on the electrochemical impedance spectroscopy data obtained in step S201, for each test frequency point, the Spearman correlation coefficient between the impedance parameters (including the real part of impedance, the imaginary part of impedance, the impedance magnitude, and the impedance phase angle) and the battery state of charge is calculated. The formula for calculating the Spearman correlation coefficient is as follows:
[0046] in, Here, is the Spearman correlation coefficient; n is the sample size. ; This represents the ranking difference between the impedance parameter values and the battery state of charge values. The Spearman correlation coefficient ranges from -1 to 1, with the absolute value closer to 1 indicating a stronger correlation.
[0047] Based on the Spearman correlation coefficient, for each dimension of the impedance parameter, frequency points with high correlation to the state of charge are selected as the first characteristic frequencies. As a preferred example, the key frequency ranges selected based on the correlation strength may include: real part in f31-f39, imaginary part in f21-f29, magnitude in f31-f39, and phase angle in f21-f29. The first characteristic frequencies cover the frequency ranges highly correlated with the state of charge in the four dimensions of impedance: real part, imaginary part, magnitude, and phase angle.
[0048] Based on electrochemical impedance spectroscopy data and the first characteristic frequency, the impedance parameters corresponding to the first characteristic frequency are extracted, including the real part of the impedance, the imaginary part of the impedance, the impedance magnitude parameter, and the impedance phase angle parameter.
[0049] By using Spearman correlation coefficients to screen characteristic frequencies highly correlated with the state of charge (SPC), the impedance parameters most sensitive to SPC changes were effectively identified. Redundant frequency points with weak SPC correlation were removed, reducing the data dimensionality for subsequent processing. Simultaneously, the most representative SPC-reflecting characteristic information was retained, laying the foundation for improving SPC estimation accuracy. The measurement time for the selected characteristic frequencies was only 9.8% of the full-spectrum sweep time, significantly improving testing efficiency.
[0050] S203. Perform dimension reduction processing on the impedance parameters to obtain the dimension-reduced feature data.
[0051] Principal Component Analysis (PCA) is a data dimensionality reduction method that uses linear transformations to map original high-dimensional features to new low-dimensional features (i.e., principal components). These new features are linear combinations of the original features and are independent of each other. Principal components can preserve the main information of the original data, representing its features with as few dimensions as possible.
[0052] Principal component features refer to the new features obtained after dimensionality reduction through principal component analysis. The principal component features are orthogonal to each other and are arranged in descending order of variance contribution rate. The first principal component contains the maximum amount of information from the original data.
[0053] Principal component analysis (PCA) is performed on the impedance parameters to reduce their dimensionality, yielding the first principal component features. These first principal component features are then used as the feature data after dimensionality reduction. The specific process is as follows: Let X be the feature matrix of the state of charge (POC) prediction samples to be reduced in dimension, with dimensions m×n, where m is the number of POC features and n is the number of battery samples. The min-max normalization method is used to process the feature matrix X, resulting in the normalized matrix X′. Normalization eliminates the influence of different dimensions and orders of magnitude, bringing all feature parameters to the same scale.
[0054] Calculate the covariance matrix C of the standardized matrix X′. The covariance matrix describes the degree of linear correlation between the characteristic parameters and is the core matrix of principal component analysis. The formula for calculating the covariance matrix C is as follows:
[0055] In the formula, is the standardized feature matrix, and m is the number of charged state features.
[0056] Find the eigenvalues λ and corresponding eigenvectors u of the covariance matrix C. The eigenvalues λ reflect the amount of information contained in the corresponding principal components, and the eigenvectors u determine the direction of the principal components. The eigenvalues and eigenvectors satisfy the following relationship:
[0057] in, For the j-th eigenvector, Let be the j-th eigenvalue.
[0058] Arrange the eigenvalues in descending order as follows: λ1≥λ2≥…≥λ m The corresponding feature vectors are u1, u2, ..., u m j = 1, 2, ..., m.
[0059] Next, the variance contribution rate and cumulative variance contribution rate of each principal component are calculated, and the variance contribution rate η of the k1th principal component is calculated. k1 This indicates the proportion of information contained in the principal component relative to the total information in the original data. The calculation formula is as follows:
[0060] in, The variance contribution rate of the k1th principal component. It is the k1th eigenvalue.
[0061] The cumulative variance contribution rate of the first p principal components represents the proportion of the sum of information contained in the first p principal components to the total information content of the original data. The calculation formula is as follows:
[0062] in, This represents the cumulative variance contribution rate of the first p principal components.
[0063] Finally, the eigenvector matrix is constructed, and the dimensionality-reduced matrix is obtained. As a preferred example, refer to... Figure 3 The impedance imaginary part PCA variance interpretation diagram shown shows that the cumulative variance contribution rate of the first two principal components (PC1 and PC2) in each dimension exceeds 99%, indicating that the first two principal components can fully represent the main information of the original impedance parameters. The first p principal components (p=2 in this embodiment) whose cumulative variance contribution rate reaches a preset threshold are selected to construct the eigenvector matrix U=[u1,u2,…,u…]. p Projecting the standardized feature matrix X′ onto the eigenvector matrix U yields the dimension-reduced matrix H:
[0064] Each column of the dimension-reduced matrix H is taken as the first principal component feature, which is the feature data after dimension reduction.
[0065] Principal component analysis (PCA) was used to reduce the dimensionality of the selected impedance parameters, resolving the potential multicollinearity problem among feature frequency points, eliminating feature redundancy, and generating independent principal component features. The dimensionality-reduced feature data significantly reduced the input dimensionality while retaining the main information of the original impedance parameters, alleviating the burden of subsequent model training, and avoiding the curse of dimensionality and overfitting risks. This creates favorable conditions for building an efficient and robust state-of-charge estimation model.
[0066] S204. Based on the dimensionality-reduced feature data, construct a charge state estimation model.
[0067] Gaussian Process Regression (GPR) is a nonparametric probabilistic model based on Bayesian theory used to establish a mapping relationship between input and output variables. GPR assumes that the output variable follows a Gaussian process distribution, learns the functional relationship between input and output through training data, and provides an estimate of the uncertainty of the prediction results.
[0068] The kernel function, also known as the covariance function, is a core component of Gaussian process regression. It characterizes the similarity between any two samples in the input space and determines the model's ability to represent data features. Commonly used kernel functions include the radial basis function (RBF) kernel and the Matrn kernel.
[0069] Hyperparameters are parameters that need to be pre-defined in a Gaussian process regression model, including the length scaling parameter of the kernel function, the signal variance parameter, and the noise variance parameter. The values of hyperparameters directly affect the model's predictive performance and generalization ability.
[0070] Exponential-Trigonometric Optimization (ETO) is a novel metaheuristic optimization algorithm that constructs adaptive dynamic parameters through the deep fusion of exponential and trigonometric functions. Combined with a two-stage exploration / exploitation mechanism and a constrained exploration strategy, it achieves a precise balance between global exploration and local exploitation, and is suitable for global optimization problems in continuous, high-dimensional parameter spaces.
[0071] Using the dimensionality-reduced feature data, battery temperature data, and battery aging state data as input, and the state of charge (SOC) of the lithium-ion battery as output, an initial estimation model is constructed. An optimization algorithm is then used to globally optimize the hyperparameters of the initial estimation model to obtain the first hyperparameter combination. Based on this first hyperparameter combination, the SOC estimation model is trained. The specific process is as follows: Using the dimensionality-reduced feature data obtained in step S203, the temperature data of the lithium-ion battery, and the aging state data of the lithium-ion battery as input variables, and the state of charge of the lithium-ion battery as the output variable, a Gaussian process regression initial estimation model is constructed. The preprocessed feature data is randomly divided into training and test sets in a 7:3 ratio.
[0072] The mathematical basis of Gaussian process regression stems from the prior assumptions of Gaussian processes: for input variables... (w is the input dimension, and R represents the set of all real numbers,) (representing w-dimensional real space), and its corresponding output variable y follows a Gaussian process distribution, i.e.:
[0073] in, This is the mean function of the Gaussian process, which is usually simplified to 0 in engineering applications to reduce computational complexity; The kernel function (covariance function) is primarily used to characterize any two samples in the input space. and The similarity between them determines the model's ability to characterize data features; This is the core set of hyperparameters for GPR. The length scaling parameter of the kernel function controls the range and sensitivity of the influence of each input variable on the output result; The signal variance parameter controls the overall fluctuation range of the output result. The noise variance parameter determines the model's tolerance to noise components in the data.
[0074] In engineering practice, the most commonly used kernel function is the radial basis function (RBF kernel), which has the following specific form:
[0075] In the formula, For the Dirac function, when The value is 1 when noise is introduced, and 0 otherwise, to adapt to noise interference in the actual data.
[0076] For a given training set GPR obtains test samples through Bayesian inference. The posterior predicted distribution:
[0077] Where: posterior mean , This is the covariance matrix of the training set (the elements are calculated using the kernel function). The covariance vector of the test sample and all training samples. Output vector for the training set; posterior variance Used to quantify the uncertainty of prediction results.
[0078] As can be seen from the core formulas above, the predictive performance of the GPR model (including prediction accuracy and uncertainty quantification accuracy) is entirely determined by the hyperparameters. The value of the hyperparameters is determined by the fact that proper configuration of the hyperparameters is a prerequisite for fully leveraging the advantages of the GPR model.
[0079] Traditional hyperparameter optimization methods (such as grid search and random search) suffer from low efficiency and are prone to getting trapped in local optima, while existing metaheuristic algorithms still fall short in balancing global exploration and local exploitation capabilities.
[0080] The purpose of this application is to overcome the shortcomings of the prior art and provide a GPR hyperparameter optimization method based on the ETO algorithm. Through the two-stage exploration / exploration mechanism and restricted exploration strategy of ETO, the method achieves efficient and global optimization of GPR hyperparameters and improves the model prediction performance.
[0081] To achieve the above objectives, this application provides a hyperparameter optimization method for Gaussian process regression based on the exponential triangular optimization algorithm, such as... Figure 4 As shown, this application provides a flowchart of a hyperparameter optimization method for Gaussian process regression using the exponential triangular optimization algorithm. This method specifically includes the following steps: Step S2041: Initialization phase.
[0082] Initialize and configure the exponential triangular optimization algorithm for hyperparameters of the Gaussian process regression model. First, define the core operating parameters of the algorithm, setting the population size to a range of 20 to 50 and the maximum number of iterations T. max The value range is set to 100-1000, and the stage switching threshold is set accordingly. Set the search boundaries for GPR hyperparameters: Low (lower bound) and Up (upper bound), depending on the kernel type, such as the RBF kernel. ∈[0.01,10]、 ∈[0.001,1]、 ∈[0.001,1], and fix the ETO core coefficients a=4.6 and b=1.55; where the stage switching threshold is calculated by the following formula:
[0083] In the formula This is the floor function.
[0084] After configuring the basic algorithm parameters, hyperparameter encoding for Gaussian process regression is performed. This involves mapping the set of hyperparameters to be optimized in the model to the dimensions of the candidate solution vectors of the exponential triangular optimization algorithm. The elements of the candidate solution vectors completely match the components of the hyperparameter set, thus achieving effective encoding of hyperparameters in the optimization algorithm. Specifically, the candidate solution vectors... The elements and the set of hyperparameters to be optimized in the GPR model The components are completely identical, that is yes The encoding form and formula in the ETO algorithm are as follows:
[0085] In the formula The dimension of the candidate solution vector (i.e., the total number of hyperparameters).
[0086] Within the preset search boundaries Within this process, N candidate solutions are generated through random sampling to form the initial population for the ETO algorithm. The generation formula is as follows:
[0087] In the formula Indicates the first iteration when the... There are 10 candidate solutions. For generated A uniformly distributed random vector in dimension [0,1], where · represents element-wise multiplication.
[0088] After the initial population is generated, the fitness of each candidate solution within the population is evaluated. The hyperparameter configurations corresponding to the candidate solutions are then substituted into the Gaussian process regression model for training. The root mean square error of the model's predictions is calculated using the five-fold cross-validation method. This error value is used as the fitness value of the corresponding candidate solution. The smaller the fitness value, the better the corresponding hyperparameter configuration. The fitness calculation formula is as follows:
[0089] In the formula, For the verification set The true value of each sample For the GPR model in the first Predicted values for each sample This represents the total number of samples in the validation set.
[0090] The candidate solution with the smallest fitness value is selected from the initial population and chosen as the optimal solution for the current iteration. And record its corresponding minimum fitness value. The initialization formula is as follows:
[0091]
[0092] Simultaneously, initialize the iteration count t=1. After initializing the optimal solution, set the initial iteration count to t=1 to prepare for subsequent iteration optimization.
[0093] Step S2042: Iteratively optimize the main loop.
[0094] The iterative optimization main loop continues to execute until the current iteration count reaches the preset maximum iteration count t=T. max First, a constrained exploration strategy is executed, and the constraint exploration iteration trigger point is calculated. This trigger point serves as the basis for determining the search boundary for dynamically updating hyperparameters. The calculation formula is as follows: (until) ) in, To constrain the exploration iteration trigger point, Let i be an exponential function with base b, where b is the core coefficient of the first algorithm, and i2 is the index. This represents the maximum number of iterations.
[0095] Recalculate the first dynamic constraint parameters and the first dynamic constraint parameters This is used to adjust the shrinkage of the search boundary and the update step size of the candidate solution. The calculation formula is as follows:
[0096]
[0097] Based on the comparison between the current iteration count and the constraint exploration iteration trigger point, determine whether to update the search boundary. Then, based on the current optimal solution The hyperparameter search boundary is dynamically updated to narrow the invalid search range. The update formula is as follows:
[0098]
[0099] In the formula, This is the updated upper limit of the search boundary. This is the updated lower bound of the search boundary. The first uniformly distributed random number with values in the range [0,1] is used. The values are random numbers from the second uniform distribution, ranging from [0,1]. This is the suboptimal solution for the current iteration; if If so, the original search boundary remains unchanged.
[0100] After adjusting the search boundary, the switching coefficient is calculated by combining uniformly distributed random numbers in the [0,1] range with the tangent function. This is used to determine the current iteration's running mode (exploration mode or development mode), and the calculation formula is as follows:
[0101] In the formula The values are random numbers from the third uniform distribution, with a range of [0,1]. It is the tangent function.
[0102] Based on the relationship between the current iteration number t and the stage switching threshold T, the iteration process is divided into two stages to achieve dynamic adaptation of exploration and development intensity. If t ≤ T, the process enters the first stage, i.e., the early iteration stage. This stage focuses on global exploration while also considering local development. The core objective is to fully expand the hyperparameter search space and discover more potential optimal solution regions. If t > T, the process enters the second stage, i.e., the later iteration stage. This stage focuses on local development while retaining a small amount of global exploration. The core objective is to conduct a refined search within the neighborhood of the discovered potential optimal solutions to improve the accuracy of hyperparameter configuration.
[0103] Based on the switching coefficient The value result is used to switch to the corresponding running mode, and a new generation of candidate solutions is generated by the position update formula of the exponential trigonometric optimization algorithm.
[0104] when When entering exploration mode, first calculate the dynamic weight coefficient. The range of candidate solutions is adjusted using the following formula:
[0105] based on The position of the candidate solution is updated using the following formula:
[0106] In the formula The values are random numbers from the fourth uniform distribution, with a range of [0,1]. For the first During the nth iteration There are 10 candidate solutions. For the first During the nth iteration There are 10 candidate solutions.
[0107] In the second phase of exploration mode, the dynamic weighting coefficients are recalculated. This guides candidate solutions to break away from the current optimal solution and explore independently. The calculation formula is as follows:
[0108] based on The position of the candidate solution is updated using the following formula:
[0109] In the formula It is a fifth uniformly distributed random number with values ranging from [0,1].
[0110] when When entering development mode, and in the first stage of development mode, calculate the dynamic weighting coefficient. The local development radius of candidate solutions is adjusted using the following formula:
[0111] based on The position of the candidate solution is updated using the following formula:
[0112] In the formula The value is a sixth uniformly distributed random number in the range [0,1]. The values are random numbers from the seventh uniform distribution, with a range of [0,1]. This is the historical best value of the i3rd candidate solution at the t-th iteration.
[0113] In the second stage of the development mode, the boundary policy coefficient c(t) is calculated to maintain the diversity of candidate solutions and avoid population homogenization. The calculation formula is as follows:
[0114] Reuse and combined The position of the candidate solution is updated using the following formula:
[0115] After the generation of new generation candidate solutions is completed, their fitness is re-evaluated. The fitness evaluation method from the initialization phase is reused to calculate the fitness value of each new generation candidate solution. The calculation formula is as follows:
[0116] Next, compare the fitness values of the new generation of candidate solutions with the current minimum fitness value. If there is a solution with a smaller fitness value among the new generation of candidate solutions, update that solution as the current best solution, and simultaneously update the current minimum fitness value. The update formula is as follows:
[0117]
[0118] If no solution exists, the current optimal solution and minimum fitness value remain unchanged. Finally, increment the current iteration count, setting t = t + 1, and return to the constrained exploration strategy execution phase to continue the iterative optimization process until the termination condition is met.
[0119] Step S2043: Termination and model application.
[0120] When the current iteration number At this point, the hyperparameter optimization process of the exponential triangular optimization algorithm officially terminates, and the final optimal combination of hyperparameters obtained in this optimization is output. This combination represents the optimal hyperparameter configuration for the Gaussian process regression model.
[0121] The final optimal hyperparameter combination Substituting the Gaussian process regression initial estimation model, the dimensionality-reduced feature data, battery temperature data, and battery aging state data were used as model inputs, and the state of charge of the lithium-ion battery was used as the model output. The model was then fully trained using all the training data. After training, the root mean square error and coefficient of determination were used to evaluate the model's performance, referring to... Figure 5 The SOC estimation results shown in the figure demonstrate that the estimated values output by the ETO-GPR model constructed in this application are in high agreement with the true values, verifying the model's high accuracy and strong generalization ability. The root mean square error (RMSE) is calculated by comparing the true values of the test set samples with the model's predicted values; a smaller value indicates higher prediction accuracy. The calculation formula is as follows:
[0122]
[0123] In the formula For the test set The true value of each sample This is the final predicted value from the GPR model. The total number of samples in the test set. The mean of the true values in the test set. The coefficient of determination (the closer it is to 1, the better the model fit).
[0124] Through the above hyperparameter optimization and model training process, the state of charge estimation model is constructed. This model can accurately output the state of charge of lithium-ion batteries based on the input feature data.
[0125] S205. Obtain the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation.
[0126] Under static or dynamic operating conditions, multi-frequency AC excitation signals are applied to the lithium-ion battery; the response data of the lithium-ion battery to the multi-frequency AC excitation signals is collected; based on the response data, the first multi-frequency electrochemical impedance spectroscopy data is calculated. The specific process is as follows: First, a set of preset multi-frequency AC excitation signals is applied to the lithium-ion battery through the AC excitation signal generation unit in the impedance-state-of-charge monitoring device. These multi-frequency AC excitation signals cover a characteristic frequency range related to the state of charge, including multiple frequency points such as low-frequency, mid-frequency, and high-frequency bands, to ensure comprehensive excitation of the battery's impedance response in different electrochemical processes. The amplitude of the multi-frequency AC excitation signals is controlled within a small signal range to avoid interfering with the battery's current state of charge.
[0127] Secondly, the voltage and current response data of the lithium-ion battery to the multi-frequency AC excitation signal are acquired in real time using the response signal acquisition unit. During the acquisition process, the surface temperature of the battery, the current operating condition (such as constant current discharge, pulse charge and discharge, rest, etc.), and timestamp information are recorded simultaneously to facilitate subsequent operating condition correlation analysis of the impedance spectrum data.
[0128] Next, based on the acquired response data, the impedance calculation unit uses a fast Fourier transform or orthogonal correlation demodulation algorithm to calculate the real part, imaginary part, magnitude, and phase angle of the impedance at each frequency point, forming the first multi-frequency electrochemical impedance spectroscopy data. This data reflects the electrochemical response characteristics of the lithium-ion battery to AC excitation at different frequencies under the current operating conditions.
[0129] Through the above steps, this application enables the rapid acquisition of multi-frequency electrochemical impedance spectroscopy data during the actual operation of lithium-ion batteries without interrupting their working state. This overcomes the limitations of traditional frequency sweeping methods, such as long testing time and inability to be applied online, and provides data support for subsequent real-time estimation of the state of charge.
[0130] S206. Based on the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model, the target state of charge of the lithium-ion battery is obtained.
[0131] From the first multi-frequency electrochemical impedance spectroscopy data, a first impedance parameter corresponding to a first characteristic frequency is extracted; the extracted first impedance parameter is subjected to the aforementioned dimensionality reduction processing to obtain first dimensionality-reduced feature data; the first dimensionality-reduced feature data is input into the state of charge estimation model to output the target state of charge of the lithium-ion battery. The specific process is as follows: First, based on the first characteristic frequency selected in step S202, impedance parameters at the corresponding frequency points are extracted from the first multi-frequency electrochemical impedance spectroscopy data. These parameters include the real part of the impedance, the imaginary part of the impedance, the impedance magnitude, and the impedance phase angle, forming a first set of impedance parameters. This set retains characteristic information highly correlated with the state of charge and removes the influence of redundant frequency points.
[0132] Secondly, the principal component analysis dimensionality reduction model established in step S203 is used to perform the same dimensionality reduction processing on the extracted first impedance parameter. Specifically, using pre-saved standardized parameters and eigenvector matrices, the first impedance parameter is mapped to the principal component space to obtain real-time dimensionality-reduced feature data isomorphic to the training phase, i.e., the first dimensionality-reduced feature data. This data significantly reduces the input dimensionality while retaining the original impedance information, ensuring the consistency of the model input.
[0133] Next, the first dimensionality-reduced feature data, along with the currently collected battery temperature data and battery aging state data, are input into the state of charge estimation model trained in step S204. The model, based on a Gaussian process regression algorithm and combined with optimized hyperparameter combinations, performs nonlinear mapping and probabilistic inference on the input features, outputting the target state of charge value of the lithium-ion battery at the current moment. Optionally, it provides the uncertainty interval of this estimate to enhance the reliability of the estimation results.
[0134] Finally, the estimated target state of charge is sent to the vehicle controller or battery management system for battery energy management, range calculation, and charge / discharge strategy optimization. The result can also be displayed in real-time on the instrument panel or a host computer interface for reference by users or maintenance personnel.
[0135] Through the above steps, this application achieves high-precision real-time monitoring of the state of charge of lithium-ion batteries based on online multi-frequency electrochemical impedance spectroscopy. It is adaptable to dynamic operating conditions, and the estimation process does not require interruption of battery operation, thus meeting the real-time and reliability requirements of online battery status monitoring in applications such as electric vehicles.
[0136] The above text combined Figures 1 to 5 The method for monitoring the state of charge of lithium-ion batteries provided in the embodiments of this application has been described in detail. The apparatus and equipment provided in the embodiments of this application will be described below with reference to the accompanying drawings.
[0137] This application also provides a device for monitoring the state of charge of a lithium-ion battery, such as... Figure 6 As shown in the figure, this is a schematic diagram of a lithium-ion battery state-of-charge monitoring device provided in an embodiment of this application. The device includes: The acquisition module 301 is used to acquire electrochemical impedance spectroscopy data of lithium-ion batteries at different temperatures, different aging states, and different states of charge; based on the electrochemical impedance spectroscopy data, the impedance parameters at characteristic frequencies related to the state of charge are extracted; and the impedance parameters are subjected to dimensionality reduction processing to obtain dimensionality-reduced feature data. The calculation module 302 is used to construct a state of charge estimation model based on the dimensionality-reduced feature data; acquire the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation; and obtain the target state of charge of the lithium-ion battery according to the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model.
[0138] In some possible implementations, the acquisition module 301 is specifically used to calculate the correlation coefficient between impedance parameters and state of charge at different frequencies based on the electrochemical impedance spectroscopy data; to select a first characteristic frequency that is highly correlated with the state of charge based on the magnitude of the correlation coefficient; and to extract the impedance parameters at the first characteristic frequency based on the electrochemical impedance spectroscopy data and the first characteristic frequency.
[0139] In some possible implementations, the acquisition module 301 is specifically used to perform principal component analysis to reduce the dimensionality of the impedance parameters and obtain the first principal component features; and to use the first principal component features as the feature data after dimensionality reduction.
[0140] In some possible implementations, the calculation module 302 is specifically used to take the dimensionality-reduced feature data, battery temperature data, and battery aging state data as inputs, and the state of charge of the lithium-ion battery as output to construct an initial estimation model; to use an optimization algorithm to globally optimize the hyperparameters of the initial estimation model to obtain a first hyperparameter combination; and to train a state of charge estimation model based on the first hyperparameter combination.
[0141] In some possible implementations, the calculation module 302 is specifically used to apply a multi-frequency AC excitation signal to the lithium-ion battery when the lithium-ion battery is in a static or dynamic operating condition; collect the response data of the lithium-ion battery to the multi-frequency AC excitation signal; and calculate the first multi-frequency electrochemical impedance spectroscopy data based on the response data.
[0142] In some possible implementations, the calculation module 302 is specifically used to extract a first impedance parameter corresponding to the first characteristic frequency from the first multi-frequency electrochemical impedance spectroscopy data; perform the dimensionality reduction processing on the extracted first impedance parameter to obtain first dimensionality-reduced feature data; input the first dimensionality-reduced feature data into the state of charge estimation model, and output the target state of charge of the lithium-ion battery.
[0143] In some possible implementations, the apparatus further includes, before acquiring electrochemical impedance spectroscopy data of the lithium-ion battery at different temperatures, aging states, and states of charge: The determination module is used to perform standard charge-discharge cycles on lithium-ion batteries to determine the actual discharge capacity of the lithium-ion batteries.
[0144] The lithium-ion battery state-of-charge monitoring device according to the embodiments of this application can correspond to the execution of the method described in the embodiments of this application, and the other operations and / or functions of each module / unit of the lithium-ion battery state-of-charge monitoring device are respectively for implementing Figure 1 For the sake of brevity, the corresponding processes of each method in the illustrated embodiments will not be described in detail here.
[0145] This application also provides a computing device. For example... Figure 7 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 400 includes a bus 401, a processor 402, a communication interface 403, and a memory 404. The processor 402, the memory 404, and the communication interface 403 communicate with each other via the bus 401.
[0146] Bus 401 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0147] Processor 402 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).
[0148] Communication interface 403 is used for communication with external devices.
[0149] Memory 404 may include volatile memory, such as random access memory (RAM). Memory 404 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).
[0150] The memory 404 stores executable code, and the processor 402 executes the executable code to perform the aforementioned method for monitoring the state of charge of a lithium-ion battery.
[0151] Specifically, in achieving Figure 6 In the case of the illustrated embodiment, and Figure 6 When the modules or units of the lithium-ion battery state-of-charge monitoring device described in the embodiments are implemented by software, the following steps are performed: Figure 6 The software or program code required for the functions of each module / unit can be partially or wholly stored in the memory 404. The processor 402 executes the program code corresponding to each unit stored in the memory 404 to execute the aforementioned method for monitoring the state of charge of the lithium-ion battery.
[0152] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the above-described method for monitoring the state of charge of a lithium-ion battery.
[0153] This application also provides a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated.
[0154] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired (e.g., coaxial cable, fiber optic) or wireless (e.g., infrared, wireless, microwave, etc.) means.
[0155] When the computer program product is executed by a computer, the computer performs any of the aforementioned methods for monitoring the state of charge of a lithium-ion battery. The computer program product can be a software installation package; when any of the aforementioned methods for monitoring the state of charge of a lithium-ion battery is required, the computer program product can be downloaded and executed on the computer.
[0156] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.
[0157] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be covered within the scope of protection of this application.
Claims
1. A method for monitoring the state of charge of a lithium-ion battery, characterized in that, The method includes: To obtain electrochemical impedance spectroscopy data of lithium-ion batteries under different temperatures, aging conditions, and charging states; Based on the electrochemical impedance spectroscopy data, impedance parameters at characteristic frequencies related to the state of charge are extracted; The impedance parameters are subjected to dimensionality reduction processing to obtain the dimensionality-reduced feature data; Based on the reduced-dimensional feature data, a state of charge estimation model is constructed; Acquire the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation; Based on the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model, the target state of charge of the lithium-ion battery is obtained.
2. The method according to claim 1, characterized in that, The extraction of impedance parameters at characteristic frequencies related to the state of charge based on the electrochemical impedance spectroscopy data includes: Based on the electrochemical impedance spectroscopy data, the correlation coefficients between impedance parameters and state of charge at different frequencies were calculated. Based on the magnitude of the correlation coefficient, the first characteristic frequency that has a high correlation with the state of charge is selected; Based on the electrochemical impedance spectroscopy data and the first characteristic frequency, the impedance parameter at the first characteristic frequency is extracted.
3. The method according to claim 1, characterized in that, The step of performing dimensionality reduction processing on the impedance parameters to obtain dimensionality-reduced feature data includes: Principal component analysis was performed on the impedance parameters to reduce their dimensionality and obtain the first principal component features. The first principal component features are used as the feature data after dimensionality reduction.
4. The method according to claim 1, characterized in that, The construction of the state of charge estimation model based on the dimensionality-reduced feature data includes: The reduced feature data, battery temperature data, and battery aging state data are used as inputs, and the state of charge of the lithium-ion battery is used as the output to construct an initial estimation model. An optimization algorithm is used to globally optimize the hyperparameters of the initial estimation model to obtain the first hyperparameter combination; Based on the first hyperparameter combination, a state-of-charge estimation model is trained.
5. The method according to claim 1, characterized in that, The acquisition of the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation includes: Apply multi-frequency AC excitation signals to the lithium-ion battery under static or dynamic operating conditions. Collect response data of the lithium-ion battery to the multi-frequency AC excitation signal; Based on the response data, the first multi-frequency electrochemical impedance spectroscopy data is calculated.
6. The method according to claim 1, characterized in that, The step of obtaining the target state of charge of the lithium-ion battery based on the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model includes: Extract the first impedance parameter corresponding to the first characteristic frequency from the first multi-frequency electrochemical impedance spectroscopy data; The extracted first impedance parameter is subjected to the aforementioned dimensionality reduction processing to obtain the first dimensionality-reduced feature data; The first dimensionality reduction feature data is input into the state of charge estimation model, and the target state of charge of the lithium-ion battery is output.
7. The method according to claim 1, characterized in that, Before acquiring electrochemical impedance spectroscopy data of lithium-ion batteries at different temperatures, aging states, and charging states, the method further includes: Perform standard charge-discharge cycles on the lithium-ion battery to determine its actual discharge capacity.
8. A device for monitoring the state of charge of a lithium-ion battery, characterized in that, The device includes: The acquisition module is used to acquire electrochemical impedance spectroscopy data of lithium-ion batteries at different temperatures, different aging states, and different states of charge; based on the electrochemical impedance spectroscopy data, impedance parameters at characteristic frequencies related to the state of charge are extracted; and the impedance parameters are subjected to dimensionality reduction processing to obtain dimensionality-reduced feature data. The calculation module is used to construct a state of charge estimation model based on the dimensionality-reduced feature data; acquire the first multi-frequency electrochemical impedance spectroscopy data of the lithium-ion battery during operation; and obtain the target state of charge of the lithium-ion battery according to the first multi-frequency electrochemical impedance spectroscopy data and the state of charge estimation model.
9. A computing device, characterized in that, Including memory and processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method as described in any one of claims 1 to 7.