Battery soh accurate estimation method based on energy storage battery attenuation characteristics and xgboost algorithm

Abstract

The present application relates to a kind of battery SOH accurate estimation method based on energy storage battery attenuation characteristic and XGBoost algorithm.It is suitable for the technical field of health management of retired energy storage battery.The technical scheme used in the present application is: the battery SOH accurate estimation method includes: obtaining the voltage, current and temperature variation data corresponding to the current charging segment of the battery to be estimated;From the voltage, current and temperature variation data, extract the multi-class battery attenuation characteristics of the current charging segment;The battery SOH estimation model trained by the battery SOH estimation model training method is input into the battery SOH estimation model, and the battery SOH of the battery to be estimated is obtained;The battery SOH estimation model training method includes: obtaining the voltage, current and temperature variation data corresponding to the battery of multiple charging segments respectively, and the battery SOH corresponding to each charging segment;From the voltage, current and temperature variation data corresponding to the charging segment, extract multiple types of battery attenuation characteristics.

Description

Technical Field

[0001] This invention relates to a method for accurately estimating the state of health (SOH) of a battery based on the degradation characteristics of energy storage batteries and the XGBoost algorithm. It is applicable to the field of health management technology for retired energy storage batteries. Background Technology

[0002] The rapid development of my country's industry has led to a continuous increase in demand for various energy sources. Simultaneously, advancements in energy storage equipment have enabled the storage of some new energy sources (such as wind power and photovoltaics) in the form of electricity. Energy storage batteries, due to their high energy density and stability, have been widely used in practice. Similarly, this has resulted in a large number of retired energy storage batteries. Retired energy storage batteries refer to batteries that have been replaced after a period of use in energy storage systems, but they still possess considerable economic value. For example, when the state-of-health (SOH) of an energy storage battery reaches 80%, although it cannot be used in electric vehicles, it can still be used in household appliances, power tools, and other devices. However, the safety of the secondary use of retired energy storage batteries remains a very important issue.

[0003] To ensure the proper recycling and reuse of retired energy storage batteries, health management technologies for these batteries are continuously evolving. Real-time monitoring and diagnostics of retired energy storage batteries can predict their lifespan and health status. State of Health (SOH) is a crucial indicator for assessing the health of retired energy storage batteries, providing key information for reuse and evaluating reusability, performance degradation, and safety during secondary use or recycling. Accurately estimating the SOH of retired batteries helps decision-makers make informed choices regarding battery reuse, recycling, or disposal. Therefore, research on SOH estimation for retired energy storage batteries is necessary.

[0004] Due to the numerous factors and complex mechanisms affecting the state-of-the-art (SOH) degradation of retired energy storage batteries, accurate estimation of SOH has always been a pressing problem for the industry. To obtain accurate SOH estimates for retired energy storage batteries, research teams both domestically and internationally have proposed many methods, including direct measurement algorithms, model-based methods, and data-driven methods.

[0005] The direct measurement method refers to obtaining battery health information and estimating SOH (State of Health) by conducting physicochemical performance tests on retired energy storage batteries in the laboratory or in actual operation, such as electrochemical impedance spectroscopy, charge-discharge capacity, and cycle life. The advantages of this method are its directness and reliability; however, it requires a lengthy experimental process and a large amount of test data, and the accuracy of the results is difficult to guarantee when the internal lifespan degradation mechanism of the battery is uncertain.

[0006] Model-based methods predict battery life (SOH) by establishing mathematical models to simulate the charging and discharging process and internal lifespan degradation mechanisms. This approach can predict battery life by simulating various charge-discharge cycles and different temperatures, and offers considerable freedom in model construction and parameter determination. However, the accuracy of the model is limited by its assumptions and the precision of the parameter determination, and requires significant time and manpower for model development and debugging.

[0007] Data-driven approaches predict battery state of health (SOH) by monitoring and analyzing actual battery operating data and utilizing technologies such as machine learning and artificial neural networks. This method requires a large amount of operational data and excellent algorithms, but does not require prior knowledge of the battery's internal mechanisms, making it flexible and easy to implement. However, the quality and quantity of data significantly impact the accuracy of the results, necessitating effective feature selection and model optimization. Summary of the Invention

[0008] The technical problem to be solved by this invention is to provide a method for accurate estimation of battery SOH based on the degradation characteristics of energy storage batteries and the XGBoost algorithm, in order to address the above-mentioned problems.

[0009] The technical solution adopted in this invention is: a method for accurate estimation of battery SOH based on energy storage battery degradation characteristics and XGBoost algorithm, characterized by comprising:

[0010] Obtain the voltage, current, and temperature change data corresponding to the current charging segment of the battery to be estimated;

[0011] Extracting multiple battery degradation characteristics from voltage, current, and temperature change data for the current charging segment;

[0012] Multiple battery degradation characteristics are input into the battery SOH estimation model trained by the battery SOH estimation model training method to obtain the battery SOH of the battery to be estimated.

[0013] The battery SOH estimation model training method includes:

[0014] Acquire voltage, current and temperature change data corresponding to multiple charging segments of the battery, as well as the battery SOH corresponding to each charging segment;

[0015] Extract various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments;

[0016] Based on the correlation between the various battery degradation characteristics of a charging segment and the state of harm (SOH) of the corresponding battery, the various battery degradation characteristics of a charging segment are screened.

[0017] Principal component analysis was used to optimize the degradation characteristics of various screened batteries.

[0018] Using the various battery degradation characteristics after screening and optimization for each charging segment as input and the battery SOH for each charging segment as output, the model is trained to obtain the mapping relationship between battery degradation characteristics and battery SOH.

[0019] The extraction of multiple battery degradation characteristics from voltage, current, and temperature change data for the current charging segment includes:

[0020] Extract various battery degradation characteristics, selected and optimized from the battery SOH estimation model training method, from the voltage, current, and temperature change data.

[0021] The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes:

[0022] Extract the constant current charging time and constant voltage charging time, as well as the ratio of constant current charging time to constant voltage charging time, from the voltage and current change data corresponding to the charging segment.

[0023] The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes:

[0024] Extract the voltage rise rate during constant current charging and the current fall rate during constant voltage charging from the voltage and current change data corresponding to the charging segment.

[0025] The rate of voltage rise during constant current charging is represented by the duration of the equal voltage interval during constant current charging; the rate of current fall during constant voltage charging is represented by the duration of the equal current interval during constant voltage charging.

[0026] The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes:

[0027] Extract the mean, variance, skewness, and kurtosis of the voltage, current, and temperature within the charging segment from the voltage, current, and temperature change data corresponding to that charging segment.

[0028] The correlation between various battery degradation characteristics based on charging segments and the state of harm (SOH) of the corresponding battery in each charging segment is used to filter the battery degradation characteristics of each charging segment, including:

[0029] The correlation between the degradation characteristics of various batteries in a charging segment and the state of harm (SOH) of the corresponding battery in that charging segment was evaluated using the Pearson coefficient and Spearman coefficient.

[0030] The process involves using the filtered and optimized battery degradation characteristics of each charging segment as input and the battery SOH of each charging segment as output to train a model, thereby obtaining the mapping relationship between battery degradation characteristics and battery SOH, including:

[0031] The XGBoost model is trained by taking the various battery degradation characteristics after screening and optimization for each charging segment as input and the battery SOH for each charging segment as output.

[0032] A battery SOH (State of Harm) accurate estimation device based on energy storage battery degradation characteristics and XGBoost algorithm, characterized in that it comprises:

[0033] The data acquisition module is used to acquire the voltage, current and temperature change data corresponding to the current charging segment of the battery to be estimated;

[0034] The feature extraction module is used to extract multiple types of battery degradation features from voltage, current, and temperature change data of the current charging segment;

[0035] The model estimation module is used to input multiple types of battery degradation features into the battery SOH estimation model trained by the battery SOH estimation model training method to obtain the battery SOH of the battery to be estimated.

[0036] The battery SOH estimation model training method includes:

[0037] Acquire voltage, current and temperature change data corresponding to multiple charging segments of the battery, as well as the battery SOH corresponding to each charging segment;

[0038] Extract various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments;

[0039] Based on the correlation between the various battery degradation characteristics of a charging segment and the state of harm (SOH) of the corresponding battery, the various battery degradation characteristics of a charging segment are screened.

[0040] Principal component analysis was used to optimize the degradation characteristics of various screened batteries.

[0041] Using the various battery degradation characteristics after screening and optimization for each charging segment as input and the battery SOH for each charging segment as output, the model is trained to obtain the mapping relationship between battery degradation characteristics and battery SOH.

[0042] The extraction of multiple battery degradation characteristics from voltage, current, and temperature change data for the current charging segment includes:

[0043] Extract various battery degradation characteristics, selected and optimized from the battery SOH estimation model training method, from the voltage, current, and temperature change data.

[0044] A storage medium storing a computer program executable by a processor, characterized in that: when the computer program is executed, it implements the steps of the battery SOH accurate estimation method based on the energy storage battery degradation characteristics and the XGBoost algorithm.

[0045] A device for accurately estimating the state of harm (SOH) of a battery, characterized in that it comprises:

[0046] processor;

[0047] The memory stores a computer program that can be executed by a processor, which, when executed, implements the steps of the battery SOH accurate estimation method based on the energy storage battery degradation characteristics and the XGBoost algorithm.

[0048] The beneficial effects of this invention are as follows: This invention extracts various types of battery degradation features from the voltage, current and temperature change data corresponding to the battery charging segment, screens degradation features that are highly correlated with SOH, and effectively eliminates redundant information through principal component analysis, thereby mining important information hidden deep in the data, providing technical support for the subsequent accurate estimation of SOH of retired energy storage batteries.

[0049] The degradation features related to SOH degradation in this invention include constant current charging time, constant voltage charging time, the ratio of constant current charging time to constant voltage charging time, the voltage rise rate during constant current charging, the current fall rate during constant voltage charging, and the mean, variance, skewness, and kurtosis of voltage, current, and temperature within the charging segment. The extracted degradation features can fully reflect the degradation characteristics of SOH, providing a data foundation for subsequent accurate estimation of SOH of retired energy storage batteries.

[0050] This invention only requires data from the current charging segment of the battery as input to the battery SOH estimation model to estimate the current SOH value of the battery, and has the advantages of high estimation accuracy and convenient operation. Attached Figure Description

[0051] Figure 1 The charging voltage and current curves for the battery charging segment.

[0052] Figure 2 The graphs show the charging voltage and current at different charging cycles.

[0053] Figure 3 A graph showing the charging voltage at different charging cycles (with an illustration of equal-interval voltages).

[0054] Figure 4 A graph showing the charging current at different charging cycles (with an illustration of equal-interval current).

[0055] Figure 5 This is a schematic diagram of principal component analysis in the embodiments.

[0056] Figure 6 The results are for verifying the estimation method in the example. Detailed Implementation

[0057] Example 1: This example is a method for training a battery SOH estimation model, which specifically includes the following steps:

[0058] S1. Obtain the voltage, current and temperature change data corresponding to each of the n charging segments of the battery, as well as the SOH of the battery corresponding to each charging segment.

[0059] In this example, battery charging is divided into two stages: constant current charging and constant voltage charging. First, the battery is charged with a constant current of 1.5A until the voltage rises to 4.2V. Then, constant voltage charging is performed until the current drops to 20mA. Figure 1 The charging voltage and current curves of battery B0006 are shown during the 20th cycle.

[0060] S2. Extract various types of battery degradation characteristics from the voltage, current, and temperature change data corresponding to the charging segment.

[0061] During battery aging, capacity can directly reflect the degree of aging, but it is not easy to measure directly. Therefore, the attenuation characteristic (AC) can be extracted from the voltage, current, and temperature change curves to indirectly reflect the aging status of the battery.

[0062] This embodiment uses voltage, current, and temperature change data corresponding to the charging segment to estimate the state of harm (SOH) of the retired energy storage battery and extracts the following features:

[0063] ① Constant current charging time and constant voltage charging time

[0064] Figure 2 This shows the changes in charging voltage and current of battery B0006 at different cycle numbers. From... Figure 2 It can be seen that the duration of constant current and constant voltage modes differs with different cycle numbers; therefore, these two time parameters are closely related to the degree of battery degradation. For this reason, the constant current charging time, constant voltage charging time, and their ratio can be extracted as battery degradation characteristics. These degradation characteristics can be calculated using equations (1)-(3):

[0065] T C =t V＝4.2 -t0 (1)

[0066] T V=t I＝0.02 -t V＝4.2 (2)

[0067] T D =T C / T V (3)

[0068] In the formula: T C T V T D These represent the constant current charging time, constant voltage charging time, and their ratio, respectively. t0 represents the start time of charging, which is usually 0. V＝4.2 t represents the time it takes for the battery voltage to reach 4.2V. I＝0.02 This indicates the time it takes for the current to drop to 0.02A.

[0069] ② Duration of the equal voltage interval in constant current mode

[0070] In section ①, the duration of constant current mode was extracted as a battery degradation feature. Further analysis... Figure 2 It was found that the rate of voltage rise during constant current charging varied with the number of cycles. For example... Figure 3 As shown, the time required for the voltage to rise from 3.8V to 4.0V varies depending on the number of cycles. Therefore, in order to more comprehensively and deeply explore the relevant features that characterize the energy storage degradation characteristics, this embodiment extracts the duration of the equal voltage interval in constant current charging mode as the degradation characteristic, as shown in equation (4):

[0071]

[0072] In the formula: This indicates the time required for the voltage to change from V1 to V2. These represent the time it takes for the voltage to reach V2 and V1, respectively.

[0073] To fully reflect the energy storage operation characteristics, this embodiment selects the middle range of voltage variation to avoid the impact of voltage fluctuations in the early and late stages on the extracted AC values. Specifically, the voltage starting point is set at 3.6V and the voltage ending point at 4V. To reduce the complexity of subsequent calculations while minimizing the loss of energy storage characteristics, a voltage interval of 0.2V is selected. That is, within the voltage range of 3.6V-4V, the duration of the voltage within the interval with a step size of 0.2V is extracted as the attenuation characteristic. The specific extracted attenuation characteristics are shown in Table 1.

[0074] Table 1. Attenuation characteristics of equal-interval voltage duration

[0075]

[0076] ③ Duration of the constant current interval under constant voltage mode

[0077] The duration of the constant voltage mode was extracted as the decay characteristic in ①, and further analysis was performed. Figure 2 It can be seen that the rate of current decrease during constant voltage charging varies with the number of cycles. For example... Figure 4 As shown, the time taken for the current to decrease from 1A to 0.5A varies under different cycles. Therefore, in order to fully characterize the charging mode of energy storage and to deeply explore the relevant features that characterize the energy storage degradation characteristics, the duration of the equal current interval under constant voltage charging mode is extracted as AC, as shown in equation (5):

[0078]

[0079] In the formula: This indicates the time required for the current to flow from C1 to C2. These represent the time it takes for the voltage to reach C2 and C1, respectively.

[0080] Similarly, in order to fully reflect the energy storage attenuation characteristics and reduce the complexity of subsequent calculations, this embodiment selects a relatively stable intermediate segment for attenuation feature extraction. The current starting point is set to 0.25A, the current ending point is set to 1.25A, and the current interval is set to 0.25A. The extracted AC is shown in Table 2.

[0081] Table 2 Attenuation characteristics of equal-interval current duration

[0082]

[0083] ④ Numerical indicators of charging segments

[0084] While complete voltage, current, and temperature data during a charging segment might adequately reflect the energy storage degradation characteristics, the sheer volume of data could lead to issues like dimensionality explosion. Therefore, to fully reflect the data characteristics, this embodiment uses mean and variance for a general description. Given a time series dataset of length L, D = [x1, x2, ..., x...],... L The mean and variance of the time series data are shown in equations (6)-(7), respectively. In addition, in order to further reflect the characteristics of the data, this embodiment also uses skewness and kurtosis to measure the symmetry and steepness of the data, as shown in equations (8)-(9).

[0085]

[0086]

[0087]

[0088]

[0089] In the formula: μ, δ, sk, and ku represent the mean, variance, skewness, and ku of the data, respectively, and L is the data length. Furthermore, when sk > 0, the time series data D is right-skewed relative to a normal distribution; when sk < 0, the time series data D is left-skewed relative to a normal distribution. When ku is between 1 and 3, the peak of the time series data D is less angular than that of a normal distribution; when ku is greater than 3, the peak of the time series data D is steeper than that of a normal distribution.

[0090] Furthermore, the mean, variance, skewness, and kurtosis of the voltage, current, and temperature of the charging segment are calculated using formulas (6)-(9). The mean, variance, skewness, and kurtosis of the voltage are denoted as μ. V δ V ,sk V , ku V The mean, variance, skewness, and kurtosis of the current are denoted as μ. C δ C ,sk C , ku C The mean, variance, skewness, and kurtosis of temperature are denoted as μ. T δ T ,sk T , ku T .

[0091] Thus, this embodiment completes the extraction of battery degradation characteristics required for SOH estimation. The extracted degradation characteristics are all derived from the voltage, current, and temperature change curves during the charging process of the energy storage battery. The extracted degradation characteristics fully characterize the main information contained in the time series data, deeply explore the potential relationship between various electrical quantities and SOH, greatly reduce the dimensionality and complexity of the data, and provide theoretical support for the subsequent accurate estimation of SOH of retired energy storage batteries.

[0092] S3. Based on the correlation between the various battery degradation characteristics of the charging segment and the SOH of the corresponding battery in the charging segment, the various battery degradation characteristics of the charging segment are screened.

[0093] This embodiment analyzes voltage, current, and temperature data during battery charging to extract battery degradation features such as constant current charging time. However, the extracted battery degradation features are numerous, and it is difficult to determine their correlation with state of charge (SOH). A large number of battery degradation features may make subsequent model training difficult. Therefore, to avoid the potential curse of dimensionality, this example filters the battery degradation features, selecting those types that have a high correlation with SOH.

[0094] In order to evaluate the correlation between the degradation characteristics of each battery and SOH, this embodiment uses the Pearson correlation coefficient and the Spearman correlation coefficient, as shown in equations (10)-(11).

[0095]

[0096]

[0097] In the formula: X represents the total sample of battery degradation characteristics; C represents the total sample of SOH; x i Represents battery degradation characteristics; c i It represents SOH.

[0098] In equations (10)-(11), the Pearson coefficient is used to evaluate the linear relationship between battery degradation characteristics and SOH, while the Spearman coefficient is used to evaluate the monotonic relationship between battery degradation characteristics and SOH. Both coefficients range from -1 to 1. The closer their absolute values ​​are to 1, the stronger the correlation; the closer they are to 0, the weaker the correlation.

[0099] Furthermore, the Pearson correlation coefficient is most appropriate when variables meet the three conditions of continuous data, normal distribution, and linear relationship. If these conditions are not met simultaneously, the Spearman coefficient is better. To fully explore the correlation between battery degradation characteristics and SOH, this embodiment combines the above two indicators for correlation evaluation. Taking the B0006 battery as an example, the correlation between the degradation characteristics of each battery and SOH is shown in Table 3.

[0100] Table 3 shows the correlation coefficients between AC and SOH.

[0101]

[0102] To comprehensively reflect the actual attenuation characteristics of SOH and control the number of input features, this embodiment selects attenuation features with both Pearson and Spearman coefficients above 0.9 as the attenuation features used for subsequent SOH estimation. The selected attenuation features are denoted as X', and the sample size is set to n; the attenuation feature matrix is ​​then denoted as X'. * The battery SOH vector under the corresponding decay characteristics is denoted as C. * .

[0103] S4. Principal component analysis is used to optimize the degradation characteristics of various batteries after screening in step S3.

[0104] Since the extracted battery degradation features are highly correlated with SOH, there may be information overlap or some potential correlations that have not been fully explored. Furthermore, due to the complexity of the SOH degradation mechanism and the large dimensionality of the input data, it is difficult to extract the key factors for estimating SOH. Therefore, to further reduce the number of data features and fully explore the potential relationships between them, this embodiment uses Principal Component Analysis (PCA) to process the battery degradation features selected in step S3. Figure 5 The diagram illustrates the dimensionality reduction of the input data after PCA processing. The specific processing steps are as follows:

[0105] 1) Solve for X * The covariance matrix COV = (X * ) T X * ;

[0106] 2) Solve for the eigenvalues ​​and eigenvectors of the covariance matrix COV, and arrange the obtained D eigenvalues ​​in descending order. Take the eigenvectors corresponding to the first d eigenvalues ​​as column vectors to form matrix P, where d <D;

[0107] 3) Transfer data X * Transform the data into a new space composed of d eigenvectors, i.e., the data after dimensionality reduction.

[0108] This completes the feature extraction of the input data. The output data is the SOH value C of the energy storage under the corresponding charging segment. * .

[0109] S5. Based on the battery degradation characteristics of each charging segment after screening and optimization in steps S3 and S4, As input, the battery SOH vector C for each charging segment * As the output, the XGBoost model is trained to obtain the mapping relationship between battery degradation characteristics and battery SOH.

[0110] This embodiment also provides a battery SOH estimation model training device, including a data acquisition module, a feature extraction module, a feature filtering module, a feature optimization module, and a model training module.

[0111] In this example, the data acquisition module acquires voltage, current, and temperature change data corresponding to multiple charging segments of the battery, as well as the state of equilibrium (SOH) of the battery for each charging segment; the feature extraction module extracts various types of battery degradation features from the voltage, current, and temperature change data corresponding to the charging segments; the feature filtering module filters various battery degradation features of the charging segments based on the correlation between the various battery degradation features of the charging segments and the SOH of the corresponding battery for that charging segment; the feature optimization module optimizes the various battery degradation features after filtering using principal component analysis; and the model training module trains the model using the filtered and optimized battery degradation features of each charging segment as input and the SOH of the battery for each charging segment as output, to obtain the mapping relationship between battery degradation features and battery SOH.

[0112] This embodiment also provides a storage medium storing a computer program that can be executed by a processor. When the computer program is executed, it implements the steps of the battery SOH estimation model training method in this example.

[0113] Example 2: This example is a method for accurately estimating the State of Harm (SOH) of a battery based on the degradation characteristics of energy storage batteries and the XGBoost algorithm. Specifically, it includes the following steps:

[0114] A. Obtain the voltage, current, and temperature change data corresponding to the current charging segment of the battery to be estimated.

[0115] B. Extract various types of battery degradation characteristics from the voltage, current, and temperature change data obtained in step A.

[0116] In this embodiment, the battery degradation feature type extracted from the voltage, current and temperature change data is the degradation feature type obtained after screening and optimization in steps S3 and S4 in Example 1.

[0117] C. Input the various battery degradation characteristics into the trained battery SOH estimation model to obtain the battery SOH of the battery to be estimated.

[0118] In this embodiment, the battery SOH estimation model is trained using the battery SOH estimation model training method in Example 1.

[0119] This embodiment also provides a battery SOH estimation device, including: a data acquisition module, a feature extraction module, and a model estimation module.

[0120] In this example, the data acquisition module is used to acquire the voltage, current, and temperature change data corresponding to the current charging segment of the battery to be estimated; the feature extraction module is used to extract multiple types of battery degradation features from the voltage, current, and temperature change data; and the model estimation module is used to input the multiple types of battery degradation features into the trained battery SOH estimation model to obtain the battery SOH of the battery to be estimated.

[0121] This embodiment also provides a storage medium storing a computer program that can be executed by a processor. When the computer program is executed, it implements the steps of the battery SOH estimation method in this example.

[0122] This embodiment also provides a battery SOH accurate estimation device, including: a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and the computer program, when executed, implements the steps of the battery SOH estimation method in this example.

[0123] When the lifespan of an energy storage battery in a new energy vehicle reaches 80%, the battery meets the requirements for retirement. Therefore, to verify the effectiveness of the SOH estimation method for retired energy storage batteries proposed in this embodiment, four batteries from the NASA dataset were selected for evaluation. The selected data includes not only data after the energy storage battery is retired, but also data from the time the energy storage battery was put into production to its retirement.

[0124] The NASA battery dataset used in this embodiment is lithium-ion battery aging data recorded by NASA during cyclic charge-discharge processes in a laboratory environment. The batteries were charged and discharged at different temperatures, and the data was recorded as charging data, discharging data, and impedance data. This embodiment selects a charging segment to estimate the state of charge (SOH) of the energy storage battery. Since the charging process in the dataset is a full-cycle charging process, this embodiment uses the ampere-hour integration method to integrate the charging segment, obtaining the SOH of the energy storage battery at the current charging cycle. This SOH value is then used as a standard value to verify the effectiveness of the method proposed in this embodiment.

[0125] To verify the results, the following indicators were set to represent the average absolute percentage error, maximum absolute percentage error, and minimum absolute percentage error of the energy storage SOH estimation results, as shown in equations (12)-(14).

[0126]

[0127]

[0128]

[0129] In the formula: y i Let be the true SOH value of the i-th sample; Let be the predicted SOH value for the i-th sample.

[0130] To match the characteristics of practical applications, this embodiment adopts a commonly used evaluation method in machine learning, namely the "leave-one-out" method. This method uses only one energy storage unit as the test set at a time, and all other energy storage units as the training set. An example setup is shown in Table 4.

[0131] Table 4. Allocation of Test and Training Sets

[0132]

[0133] The results obtained using the energy storage SOH estimation method proposed in this embodiment are as follows: Figure 6 As shown in Table 5, the values ​​of the three metrics for each test set are shown in Table 5.

[0134] Table 5. Verification results of the four energy storage batteries

[0135]

[0136] Through analysis Figure 6 Based on the estimation results of the four battery sample data in Table 5, it can be seen that the MAPE of the method proposed in this embodiment is less than 0.201% for all samples, the maximum Max_error is only 1.368%, and the minimum Min_error is as low as 0.002%. This verifies that the SOH estimation method for energy storage batteries proposed in this embodiment has high estimation accuracy. Furthermore, the results show that the method can obtain good estimation results on all four sample data, effectively verifying the robustness of the method.

[0137] To further verify that the method proposed in this embodiment still has high accuracy in estimating the SOH of the decommissioned energy storage battery, this embodiment analyzes the values ​​of various evaluation indicators for the four batteries mentioned above when the SOH of the energy storage battery reaches below 80%, that is, after the energy storage battery is decommissioned, as shown in Table 6.

[0138] Table 6. Verification results of energy storage batteries in four retirement phases.

[0139]

[0140]

[0141] Analysis of the results in Table 6 shows that the method proposed in this embodiment still has high accuracy in estimating the SOH of retired energy storage batteries. Analysis of the estimation results for four retired batteries shows that: MAPE is less than 0.216%, Max_error is only 1.238%, and Min_error is as low as 0.002%. This verifies that the SOH estimation method for retired energy storage batteries proposed in this invention can effectively estimate the SOH of retired energy storage batteries, and has high accuracy and strong robustness.

Claims

1. A method for accurately estimating the State of Harm (SOH) of a battery based on the degradation characteristics of energy storage batteries and the XGBoost algorithm, characterized in that, include: Obtain the voltage, current, and temperature change data corresponding to the current charging segment of the battery to be estimated; Extracting multiple battery degradation characteristics from voltage, current, and temperature change data for the current charging segment; Multiple battery degradation characteristics are input into the battery SOH estimation model trained by the battery SOH estimation model training method to obtain the battery SOH of the battery to be estimated. The battery SOH estimation model training method includes: Acquire voltage, current and temperature change data corresponding to multiple charging segments of the battery, as well as the battery SOH corresponding to each charging segment; Extract various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments; Based on the correlation between the various battery degradation characteristics of a charging segment and the state of harm (SOH) of the corresponding battery, the various battery degradation characteristics of a charging segment are screened. Principal component analysis was used to optimize the degradation characteristics of various screened batteries. Using the various battery degradation characteristics after screening and optimization for each charging segment as input and the battery SOH for each charging segment as output, the model is trained to obtain the mapping relationship between battery degradation characteristics and battery SOH. The extraction of multiple battery degradation characteristics from voltage, current, and temperature change data for the current charging segment includes: Extract various battery degradation characteristics, after screening and optimization, from the battery SOH estimation model training method based on voltage, current, and temperature change data; The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes: Extract the constant current charging time and constant voltage charging time, as well as the ratio of constant current charging time to constant voltage charging time, from the voltage and current change data corresponding to the charging segment. The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes: Extract the voltage rise rate during constant current charging and the current fall rate during constant voltage charging from the voltage and current change data corresponding to the charging segment. The rate of voltage rise during constant current charging is represented by the duration of the equal voltage interval in constant current mode; the rate of current fall during constant voltage charging is represented by the duration of the equal current interval in constant voltage mode. The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes: Extract the mean, variance, skewness, and kurtosis of the voltage, current, and temperature within the charging segment from the voltage, current, and temperature change data corresponding to that charging segment.

2. The method for accurate estimation of battery SOH based on energy storage battery degradation characteristics and XGBoost algorithm according to claim 1, characterized in that, The correlation between various battery degradation characteristics based on charging segments and the state of harm (SOH) of the corresponding battery in each charging segment is used to filter the battery degradation characteristics of each charging segment, including: The correlation between the degradation characteristics of various batteries in a charging segment and the state of harm (SOH) of the corresponding battery in that charging segment was evaluated using the Pearson coefficient and Spearman coefficient.

3. The method for accurate estimation of battery SOH based on energy storage battery degradation characteristics and XGBoost algorithm according to claim 1, characterized in that, The process involves using the filtered and optimized battery degradation characteristics of each charging segment as input and the battery SOH of each charging segment as output to train a model, thereby obtaining the mapping relationship between battery degradation characteristics and battery SOH, including: The XGBoost model is trained by taking the various battery degradation characteristics after screening and optimization for each charging segment as input and the battery SOH for each charging segment as output.

4. A device for accurate estimation of battery SOH based on energy storage battery degradation characteristics and XGBoost algorithm, characterized in that, include: The data acquisition module is used to acquire the voltage, current and temperature change data corresponding to the current charging segment of the battery to be estimated; The feature extraction module is used to extract multiple types of battery degradation features from voltage, current, and temperature change data of the current charging segment; The model estimation module is used to input multiple types of battery degradation features into the battery SOH estimation model trained by the battery SOH estimation model training method to obtain the battery SOH of the battery to be estimated. The battery SOH estimation model training method includes: Acquire voltage, current and temperature change data corresponding to multiple charging segments of the battery, as well as the battery SOH corresponding to each charging segment; Extract various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments; Based on the correlation between the various battery degradation characteristics of a charging segment and the state of harm (SOH) of the corresponding battery, the various battery degradation characteristics of a charging segment are screened. Principal component analysis was used to optimize the degradation characteristics of various screened batteries. Using the various battery degradation characteristics after screening and optimization for each charging segment as input and the battery SOH for each charging segment as output, the model is trained to obtain the mapping relationship between battery degradation characteristics and battery SOH. The extraction of multiple battery degradation characteristics from voltage, current, and temperature change data for the current charging segment includes: Extract various battery degradation characteristics, after screening and optimization, from the battery SOH estimation model training method based on voltage, current, and temperature change data; The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes: Extract the constant current charging time and constant voltage charging time, as well as the ratio of constant current charging time to constant voltage charging time, from the voltage and current change data corresponding to the charging segment. The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes: Extract the voltage rise rate during constant current charging and the current fall rate during constant voltage charging from the voltage and current change data corresponding to the charging segment. The rate of voltage rise during constant current charging is represented by the duration of the equal voltage interval in constant current mode; the rate of current fall during constant voltage charging is represented by the duration of the equal current interval in constant voltage mode. The extraction of various types of battery degradation characteristics from voltage, current, and temperature change data corresponding to charging segments includes: Extract the mean, variance, skewness, and kurtosis of the voltage, current, and temperature within the charging segment from the voltage, current, and temperature change data corresponding to that charging segment.

5. A storage medium storing a computer program executable by a processor, characterized in that: When the computer program is executed, it implements the steps of the battery SOH accurate estimation method based on the energy storage battery degradation characteristics and XGBoost algorithm as described in claims 1 to 3.

6. A device for accurately estimating the state of harm (SOH) of a battery, characterized in that, include: processor; The memory stores a computer program that can be executed by a processor, which, when executed, implements the steps of the battery SOH accurate estimation method based on energy storage battery degradation characteristics and XGBoost algorithm as described in claims 1-3.

Images