A lithium-ion battery capacity estimation method based on capacity increment curve characteristics

Abstract

This invention relates to a lithium-ion battery capacity estimation method based on capacity increment curve characteristics. It collects raw charge-discharge data throughout the entire lifecycle of the lithium-ion battery. Based on this data, it calculates the capacity increment IC value during the charging process, smooths it using a moving average method, calculates the first derivative of the smoothed IC curve with respect to voltage, and identifies the position of the first peak of the first derivative curve. The IC value and voltage value corresponding to this peak position are used as feature values ​​for lithium-ion battery capacity estimation. These feature values ​​are inputs, and the battery's capacity loss rate is output. A neural network model is then established to estimate the battery capacity, yielding the estimated capacity value. The lithium-ion battery capacity estimation method proposed in this invention is unaffected by aging conditions and charge-discharge rates during feature extraction, exhibits high precision and accuracy, has a simple process, and is highly systematic, making it suitable for practical estimation applications.

Description

Technical Field

[0001] This invention belongs to the field of lithium-ion battery technology, and in particular to a lithium-ion capacitance estimation method based on the characteristics of the capacity increment curve. Background Technology

[0002] Lithium-ion batteries are widely used in energy storage systems, electric vehicles, and aerospace power supplies due to their advantages such as high power density, high energy density, long cycle life, high safety, low cost, light weight, no memory effect, and environmental friendliness. However, lithium-ion batteries age over long-term use, manifesting as capacity reduction and increased impedance. Capacity reduction affects the battery's charge and discharge performance, further impacting the functionality and task completion of electrical devices. Therefore, capacity estimation of lithium-ion batteries is particularly important in battery management systems. Lithium-ion battery capacity estimation methods can be categorized into four types: experimental calibration methods, semi-empirical methods, Bayesian estimation methods, and data-driven methods.

[0003] The experimental calibration method uses a standard capacity calibration experiment to perform a complete charge and discharge experiment on the battery to obtain its capacity information. The measurement results are relatively accurate, but it is difficult to obtain complete charge and discharge experiment data in practical applications. Therefore, this method is usually used as a verification of the results of other capacity estimation methods.

[0004] Semi-empirical methods, based on extensive experiments, analyze the correlation between battery capacity decay and stress factors such as temperature and depth of discharge (DOD), establishing simple semi-empirical function models such as capacity-temperature-DOD, typically categorized into calendar aging models and cycle aging models. However, battery aging is usually influenced by multiple factors, and semi-empirical function models lack consideration for these multivariate factors, resulting in limited accuracy.

[0005] Bayesian estimation methods mainly refer to capacity estimation methods that combine Kalman filtering, particle filtering, and other algorithms with battery models. In this method, the battery capacity is considered a slowly changing state in the model, and its changes affect both the battery's open-circuit voltage and terminal voltage. Based on this, the battery's voltage error feedback can be used to update its state, thus forming a closed-loop estimation system. The disadvantages of this method are that it requires setting reasonable initial values ​​for the algorithm; otherwise, the algorithm risks divergence. Furthermore, as the battery ages, the parameters in the equivalent model will change, requiring additional parameter identification methods.

[0006] Data-driven methods treat the battery as a black box, disregarding the complex electrochemical reactions occurring within it. They directly extract feature values ​​from its charge-discharge data, using these as input to machine learning and artificial intelligence algorithms, with capacity as the output, establishing a nonlinear mapping relationship between the two. The accuracy of this method is affected by the quality of the input data; therefore, obtaining features highly correlated with battery aging from the charge-discharge data is crucial. In actual battery use, discharge is typically affected by load, and discharge data may be irregular, while the charging process is relatively controllable. Therefore, this invention provides a lithium-ion battery capacity estimation method based on the characteristics of the capacity increment curve during the charging process. Summary of the Invention

[0007] The purpose of this invention is to provide a lithium-ion capacitance estimation method based on the characteristics of the capacity increment curve.

[0008] To achieve the above objectives, this invention provides a lithium-ion battery capacity estimation method based on capacity increment curve characteristics, the specific steps of which are as follows:

[0009] (1) Collect raw charge and discharge data of lithium-ion batteries throughout their entire life cycle, including charge and discharge time, current and voltage;

[0010] (2) Based on the original charge and discharge data obtained in step (1), calculate the capacity increment (IC, Incremental capacity) value during the charging process. The calculation formula is as follows:

[0011]

[0012] Among them, IC k U represents the capacity increment at time k; k Q is the voltage value at time k; k U is the capacity value at time k. k-1 Q is the voltage value at time k-1; k-1 The capacity value at time k-1 is calculated using the ampere-hour integration method, and its discrete calculation formula is:

[0013]

[0014] Among them, Q k Let I be the capacity value at time k. k Δt represents the current value at time k; Δt is the sampling interval.

[0015] (3) Since the original charge and discharge data obtained in step (1) is affected by sampling accuracy and noise of the acquisition equipment, the IC curve of the capacity increment value obtained according to equation (1) has numerical fluctuations, which will affect the selection of feature values ​​in the IC curve. The moving average method is used to smooth it, and its calculation formula is:

[0016]

[0017] Where y is the data after smoothing by moving average; x is the original charge and discharge data; N is the smoothing window size of the moving average method; i is the current smoothing point; and j is the point traversed within the smoothing window.

[0018] (4) Calculate the first derivative of the IC curve obtained after smoothing in step (3) with respect to voltage, that is, its rate of change. The formula is:

[0019]

[0020] Wherein: dIC k U is the first derivative of the IC curve at sampling point k; k U is the voltage value at time k; k-1 The voltage value at time k-1; IC k IC represents the capacity increment at time k. k-1 This represents the capacity increment at time k-1;

[0021] (5) Calibrate the position where the first peak of the IC curve first appears, and take the peak value of the IC curve, the IC value and voltage value corresponding to the position of the first peak value as feature values ​​for lithium-ion battery capacity estimation.

[0022] (6) Using the feature values ​​obtained in step (5) as input and the battery capacity loss rate as output, a neural network model is established. The formula for calculating the capacity loss rate is:

[0023]

[0024] Among them, dQ l Q represents the capacity loss rate of the battery during the l-th cycle. ini Q represents the initial capacity of the battery. l This represents the battery capacity during the l-th cycle.

[0025] (7) Using the neural network model constructed in step (6), estimate the battery capacity. When the estimated capacity loss rate output by the model is obtained, calculate the estimated capacity value:

[0026]

[0027] in, This is the estimated capacity of the battery during the l-th cycle; This is the estimated battery capacity loss rate output by the established model during the l-th cycle.

[0028] The beneficial effects of this invention are as follows:

[0029] This invention proposes a lithium-ion battery capacity estimation method based on capacity increment curve features. It obtains the maximum value of the initial rate of change of the capacity increment curve and its corresponding voltage and capacity increment values ​​as features, and uses a neural network to construct a mapping relationship between these features and the capacity loss rate of the lithium-ion battery, thereby estimating its capacity. Due to the influence of temperature, charge / discharge rate, etc., the peaks and valleys of the capacity increment curve may disappear, making capacity estimation methods based on peak and valley features susceptible to failure. Regardless of how the capacity increment curve changes, it will always transition from a flat region to a steep region, thus the curve inevitably has a maximum rate of change. Based on this, the lithium-ion battery capacity estimation method proposed in this invention is unaffected by capacity and charge / discharge rate, has high accuracy and precision, a simple process, and high systematicity, making it suitable for practical estimation applications. Attached Figure Description

[0030] Figure 1 The IC curves of battery No. 1 in the three groups of batteries in the embodiment are shown below; (a) is the IC curve of the battery under different aging lifespans under 25℃-0.5C charge-1C discharge cycle, (b) is the IC curve of the battery under different aging lifespans under 35℃-0.5C charge-1C discharge cycle, and (c) is the IC curve of the battery under different aging lifespans under 25℃-0.25C charge-1C discharge cycle.

[0031] Figure 2 The IC rate of change curves for battery No. 1 in the three groups of batteries in the embodiment are shown below; where: (a) is the IC rate of change curve for different aging lifetimes of the battery under 25℃-0.5C charge-1C discharge cycle, (b) is the IC rate of change curve for different aging lifetimes of the battery under 35℃-0.5C charge-1C discharge cycle, and (c) is the IC rate of change curve for different aging lifetimes of the battery under 25℃-0.25C charge-1C discharge cycle.

[0032] Figure 3 The following is a schematic diagram of feature extraction in the embodiment; where: (a) is the IC curve of different aging states, (b) is the local IC curve of different aging states, (c) is the dIC curve of different aging states, and (d) is the local dIC curve of different aging states.

[0033] Figure 4 The capacity estimation results and errors of batteries #2 and #3 in group A are as follows: (a) is the capacity estimation result of battery #2 in group A, (b) is the capacity estimation error of battery #2 in group A, (c) is the capacity estimation result of battery #3 in group A, and (d) is the capacity estimation error of battery #3 in group A.

[0034] Figure 5The capacity estimation results and errors of batteries #2 and #3 in group B are as follows: (a) Capacity estimation result of battery #2 in group B, (b) Capacity estimation error of battery #2 in group B, (c) Capacity estimation result of battery #3 in group B, and (d) Capacity estimation error of battery #3 in group B.

[0035] Figure 6 The capacity estimation results and errors of batteries #2 and #3 in group C are as follows: (a) Capacity estimation result of battery #2 in group C, (b) Capacity estimation error of battery #2 in group C, (c) Capacity estimation result of battery #3 in group C, and (d) Capacity estimation error of battery #3 in group C. Detailed Implementation

[0036] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Example 1: This invention provides a lithium-ion battery capacity estimation method based on capacity increment curve characteristics, including the following steps:

[0038] Step 1: Obtain lithium-ion battery data and divide them into three groups (A, B, and C) according to their aging cycle test conditions. Group A batteries underwent aging cycle testing at 25°C with a 0.5C rate constant current constant voltage charge and a 1C rate discharge. Group B batteries underwent aging cycle testing at 25°C with a 0.25C rate constant current constant voltage charge and a 1C rate discharge. Group C batteries underwent aging cycle testing at 35°C with a 0.5C rate constant current constant voltage charge and a 1C rate discharge. The charge / discharge cutoff voltages for all batteries were 4.2V and 2.65V, respectively, and the cutoff current for the CV stage was 0.05C. Furthermore, each group of batteries contained three cells.

[0039] Step 2: Calculate the smoothed IC curve according to equations (1) to (3), as follows: Figure 1 As shown, the IC curves of the batteries under the three aging cycle tests are not entirely consistent. In group A, the first peak gradually disappears after long-term cycling; in group B, the first and third peaks gradually disappear; and in group C, the first and third peaks gradually disappear. This indicates that establishing a capacity estimation model based on peak values ​​may fail.

[0040] Step 3: Calculate the first derivative curve (i.e., its rate of change curve) of the IC curve according to equation (4), as follows: Figure 2As shown, it can be observed that the first peak changes regularly with battery aging, and its identifiability is very high.

[0041] Step 4: Based on the position of the first peak in the IC rate of change curve calibrated in Step 3, extract the maximum value of the rate of change corresponding to that peak, along with its corresponding voltage and IC values, as features. Taking battery number 1 in group A as an example... Figure 3 As shown.

[0042] Step 5: For each battery group, select the data of one battery as the training set and the remaining two batteries as the test set.

[0043] Step 6: Construct a neural network model, which consists of an input layer, a hidden layer, and an output layer. The number of neurons in the hidden layer is set to 20.

[0044] Step 7: Use the features extracted from the training set in Step 5 as input and the capacity loss rate as output to train the model established in Step 6.

[0045] Step 8: Based on the model trained in Step 7, using the features extracted from the test set in Step 5 as input, an estimated value of the capacity loss rate can be obtained, and the capacity estimate can be further calculated, such as... Figures 4 to 6 As shown.

[0046] To demonstrate the effectiveness of this invention, the estimation results are evaluated based on three error statistics: maximum absolute error (MAE), mean absolute error (MaE), and root mean square error (RMSE).

[0047] In Group A, battery #2 had a MAE of 1.14%, a MAE of 0.51%, and an RMSE of 0.54%; battery #3 had a MAE of 2.35%, a MAE of 1.22%, and an RMSE of 1.34%.

[0048] In Group B, battery #2 had a MAE of 3.51%, a MaE of 1.56%, and an RMSE of 1.79%; battery #3 had a MAE of 3.56%, a MaE of 0.82%, and an RMSE of 1.06%.

[0049] In Group C, battery #2 had a MAE of 2.06%, a MaE of 0.63%, and an RMSE of 0.75%; battery #3 had a MAE of 0.67%, a MaE of 0.29%, and an RMSE of 0.32%.

[0050] In summary, the proposed method provides a novel capacity estimation approach. Unlike traditional capacity increment peak information, the features extracted in this invention are unaffected by operating conditions and battery aging, exhibiting high identifiability. The estimation model established based on this method was validated on three sets of battery data under different test conditions. The maximum estimation error did not exceed 3.56%, the average error did not exceed 1.56%, and the root mean square error did not exceed 1.79%, demonstrating that the proposed capacity estimation method possesses high estimation accuracy, strong robustness, and versatility.

[0051] The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the concept and scope of the present invention. Without departing from the design concept of the present invention, all modifications and improvements made by those skilled in the art to the technical solutions of the present invention should fall within the protection scope of the present invention. The technical content for which protection is sought in the present invention has been fully described in the technical requirements.

Claims

1. A lithium-ion battery capacity estimation method based on capacity increment curve characteristics, characterized in that... The specific steps are as follows: (1) Collect raw charge and discharge data of lithium-ion batteries throughout their entire life cycle, including charge and discharge time, current and voltage; (2) Based on the original charge and discharge data obtained in step (1), calculate the capacity increment (IC, Incremental capacity) value during the charging process. The calculation formula is as follows: Among them, IC k U represents the capacity increment at time k; k Q is the voltage value at time k; k U is the capacity value at time k. k-1 Q is the voltage value at time k-1; k-1 The capacity value at time k-1 is calculated using the ampere-hour integration method, and its discrete calculation formula is: Among them, Q k Let I be the capacity value at time k. k Δt represents the current value at time k; Δt is the sampling interval. (3) Since the original charge and discharge data obtained in step (1) is affected by sampling accuracy and noise of the acquisition equipment, the IC curve of the capacity increment value obtained according to equation (1) has numerical fluctuations, which will affect the selection of feature values ​​in the IC curve. The moving average method is used to smooth it, and its calculation formula is: Where y is the data after smoothing by moving average; x is the original charge and discharge data; N is the smoothing window size of the moving average method; i is the current smoothing point; and j is the point traversed within the smoothing window. (4) Calculate the first derivative of the IC curve obtained after smoothing in step (3) with respect to voltage, that is, its rate of change. The formula is: Wherein: dIC k U is the first derivative of the IC curve at sampling point k; k U is the voltage value at time k; k-1 The voltage value at time k-1; IC k IC represents the capacity increment at time k. k-1 This represents the capacity increment at time k-1; (5) Calibrate the position where the first peak of the IC curve first appears, and take the peak value of the IC curve, the IC value and voltage value corresponding to the position of the first peak value as feature values ​​for lithium-ion battery capacity estimation. (6) Using the feature values ​​obtained in step (5) as input and the battery capacity loss rate as output, a neural network model is established. The formula for calculating the capacity loss rate is: Among them, dQ l Q represents the capacity loss rate of the battery during the l-th cycle. ini Q represents the initial capacity of the battery. l This represents the battery capacity during the l-th cycle. (7) Using the neural network model constructed in step (6), estimate the battery capacity. When the estimated capacity loss rate output by the model is obtained, calculate the estimated capacity value: in, This is the estimated capacity of the battery during the l-th cycle; This is the estimated battery capacity loss rate output by the established model during the l-th cycle.

Images