A method for predicting soh and rul of lithium battery

Abstract

This invention discloses a method for predicting the State of Health (SOH) and Remaining Life (RUL) of lithium batteries, relating to the field of lithium battery technology. The method includes the following steps: acquiring full-lifetime electrochemical impedance spectroscopy (EIS) data of the lithium battery; constructing an equivalent circuit model of the lithium battery; fitting the Nyquist plot in the full-lifetime EIS of the lithium battery to obtain the parameters of each component in the equivalent circuit model under different SOH states of the lithium battery; training a bagged tree algorithm using a dataset to fit the relationship between different health states, remaining lifespan, and the EIS of the lithium battery; and substituting the EIS of an unknown lithium battery into the trained bagged tree algorithm to obtain the unknown SOH and RUL of the lithium battery. This invention considers the inherent chemical characteristics of lithium batteries during the modeling process and integrates battery charge / discharge data and EIS for model parameter identification, resulting in more accurate predictions of SOH and RUL.

Description

Technical Field

[0001] This invention relates to the field of lithium battery technology, and in particular to a method for predicting the SOH and RUL of lithium batteries. Background Technology

[0002] In recent years, the increasing number of new energy vehicles, represented by electric vehicles, has led to a wave of retired lithium-ion batteries. Failure to properly manage this massive amount of lithium-ion batteries will result in serious energy waste and environmental pollution.

[0003] The secondary use of batteries can effectively alleviate the pollution problem of retired power batteries and maximize the utilization of the remaining value of lithium-ion power batteries, thereby achieving energy conservation and emission reduction. Furthermore, the secondary use of retired power batteries can also promote the development of the new energy industry. However, the testing and screening technologies for lithium-ion power batteries are relatively outdated. The currently popular method of measuring ampere-hours using full charge and full discharge methods not only has low measurement accuracy but also consumes a lot of time and manpower, and therefore does not have the prospect of large-scale application.

[0004] Currently, scholars both domestically and internationally have conducted extensive research on methods for predicting the State of Health (SOH) and Remaining Life (RUL) of lithium-ion batteries, primarily including data-driven and model-driven methods. Data-driven methods utilize historical and real-time battery data to build predictive models for SOH and RUL prediction. This approach avoids the need to construct and update complex models and offers good prediction efficiency for large-scale battery packs. However, current technologies only consider common parameters such as current, voltage, and internal resistance, neglecting the design, installation, and usage of the lithium-ion batteries themselves. Therefore, the accuracy and reliability of these predictions still need improvement. Model-driven methods first construct an internal battery model and then identify parameters to predict SOH and RUL. Current methods mainly include electrochemical models, equivalent circuit models, and empirical models. This approach is suitable for assessing the SOH of single or small numbers of lithium-ion batteries. However, for scenarios with a large number of batteries, such as electric vehicles or energy storage power stations, the workload of building models is substantial, significantly reducing the accuracy of predictions.

[0005] In summary, existing prediction methods have poor accuracy in predicting SOH and RUL. Summary of the Invention

[0006] This invention provides a method for predicting SOH and RUL of lithium batteries, which solves the problem of poor prediction accuracy of SOH and RUL in the prior art.

[0007] This invention provides a method for predicting the state of oxygen (SOH) and raw uptime (RUL) of lithium batteries, comprising the following steps:

[0008] Obtain electrochemical impedance spectroscopy data of lithium batteries throughout their entire lifespan;

[0009] Constructing an equivalent circuit model for lithium batteries based on battery electrochemistry principles;

[0010] By using the parameters of each component in the equivalent circuit model as unknowns, the Nyquist plot in the full-lifetime electrochemical impedance spectroscopy data of lithium batteries is fitted using the Levenberg-Marquardt method to obtain the parameters of each component in the equivalent circuit model under different health states (SOH) of lithium batteries, and then a dataset is constructed.

[0011] The bag tree algorithm was trained using a dataset to fit the relationship between different health states (SOH), remaining service life (RUL), and the electrochemical impedance spectroscopy of lithium batteries.

[0012] By substituting the electrochemical impedance spectroscopy data of the unknown lithium battery into the trained bagged tree algorithm, the state of health (SOH) and remaining lifetime (RUL) of the unknown lithium battery can be obtained.

[0013] Preferably, the full-lifetime electrochemical impedance spectroscopy data of lithium batteries are obtained based on multiple battery charge-discharge experiments, wherein the steps of one cycle of battery charge-discharge experiment include:

[0014] A 36A constant current discharge of a lithium battery for 5 seconds was used for the inrush current test of ohmic internal resistance.

[0015] The lithium battery is discharged at a constant current of 20A until its voltage drops below 2.65V;

[0016] Let the lithium battery stand for 15 minutes to allow the internal chemical reaction to fully occur.

[0017] Lithium battery 36A constant current charging for 5 seconds;

[0018] Lithium battery 20A constant current charging for 5 seconds;

[0019] The lithium battery is charged at a constant current of 20A until its voltage is higher than 3.7V;

[0020] Let the lithium battery stand for 15 minutes.

[0021] Preferably, the equivalent circuit model inside the lithium battery includes an ohmic internal resistance, an SEI resistor, a constant phase angle element, a charge transfer internal resistance, an electric double layer capacitor, and a Warburg impedance; the SEI resistor is connected in parallel with the constant phase angle element, and the charge transfer internal resistance is connected in series with the Warburg impedance and then connected in parallel with the electric double layer capacitor.

[0022] Preferably, the admittance expression of the constant phase angle element is as follows:

[0023] Y Q =Y O (jω)α

[0024] In the formula, Q is a constant phase angle element, Y Q For constant phase angle element admittance, Y O α is the zero-phase admittance, ω is the constant phase angle coefficient, ω is the circuit angular velocity, and j is the imaginary unit.

[0025] The expression for the Warburg impedance is as follows:

[0026]

[0027] In the formula, Z w Let σ be the Warburg impedance value, and σ be the Warburg impedance constant.

[0028] Preferably, the impedance expression of the equivalent circuit model of the lithium battery is as follows:

[0029]

[0030] In the formula, Z is the impedance of the equivalent circuit model of the lithium battery, and R... e For ohmic internal resistance, R sei For SEI resistor, C dl For double-layer capacitors, R dl This represents the internal resistance to charge transfer.

[0031] Preferably, training the bagged tree algorithm using a dataset includes the following steps:

[0032] Data sampling is performed on the dataset;

[0033] Multiple subsamples are obtained by random sampling with replacement from the dataset.

[0034] Each of the sub-samples is used to train a base model;

[0035] Multiple basic models make predictions on the input data, generating multiple prediction results;

[0036] The final prediction result is obtained by combining multiple prediction results.

[0037] Preferably, the basic model is a decision tree, a classifier, or a regression model.

[0038] Preferably, multiple prediction results are combined to obtain the final prediction result, including:

[0039] For classification problems, the category that receives the most votes is selected as the prediction result;

[0040] For regression problems, the average of the predictions from all basic models is taken as the prediction result.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] This invention acquires full-lifetime electrochemical impedance spectroscopy (EIS) data of lithium-ion batteries. Based on the electrochemical principles of batteries, an equivalent circuit model of the battery's internal structure is established, taking into account the inherent chemical characteristics of lithium-ion batteries. Then, the parameters of each component in the equivalent circuit under full-lifetime EIS are fitted. By combining battery charge / discharge data with EIS analysis, model parameters are identified, resulting in more accurate parameters. Finally, a bagged tree machine learning algorithm is used to fit the relationship between SOH, RUL, and battery ESI. This reduces variance, improves model stability, and makes the predictions of SOH and RUL more accurate. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A flowchart of a method for predicting the SOH and RUL of a lithium battery according to the present invention;

[0045] Figure 2 This is a schematic diagram of the equivalent circuit model of the lithium battery of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0047] This invention proposes a method for predicting the State of Harm (SOH) and Remaining Life (RUL) of lithium-ion batteries. First, an equivalent circuit model of the lithium-ion battery is constructed. Simultaneously, charge-discharge experiments are conducted, and the electrochemical impedance spectroscopy (EIS) of the battery at different cycle numbers is measured using an electrochemical workstation. This allows for the determination of parameters for different components within the constructed equivalent circuit model. After the solution is obtained, bag-tree processing is used for machine learning to derive the relationship between the battery's SOH, cycle number, and the parameters in the equivalent circuit. Finally, by measuring the EIS of an unknown battery using an electrochemical workstation and substituting the results of the bag-tree machine learning, the battery's SOH and remaining life can be determined. (Refer to...) Figure 1 Specifically, it includes the following steps:

[0048] Step 1: Obtain electrochemical impedance spectroscopy data of the lithium battery throughout its entire lifespan.

[0049] The data acquisition method involves designing and conducting battery charge-discharge experiments to obtain voltage, current, and electrochemical impedance spectroscopy data throughout the battery's lifespan. To obtain full-lifespan usage data, this invention first simulates battery aging and degradation (capacity decay to 80% of initial capacity) by repeatedly charging and discharging a new battery, recording voltage and current data at various points during the aging process to obtain the battery's usage history data before retirement. After completing the battery aging and retirement simulation, further charge-discharge experiments are conducted, recording voltage and current data at various points during the retirement battery's charge-discharge process.

[0050] In this embodiment, a battery charge-discharge experiment is designed using a 20Ah lithium battery as an example. The specific steps of the battery charge-discharge experiment within one cycle are as follows:

[0051] (1) Let the battery stand for 30 seconds to prepare the equipment before the experiment.

[0052] (2) 36A constant current discharge for 5s, used for the impact current test of ohmic internal resistance.

[0053] (3) Discharge at a constant current of 20A until the voltage drops below 2.65V.

[0054] (4) Let the battery stand for 15 minutes to allow the chemical reaction inside the battery to occur fully.

[0055] (5) 36A constant current charging for 5s.

[0056] (6) 20A constant current charging for 5s.

[0057] (7) Charge at a constant current of 20A until the voltage is higher than 3.7V.

[0058] (8) Let the battery stand for 15 minutes.

[0059] The above-described method can be used to design battery charge and discharge experiments for batteries ranging from 20 to 400Ah.

[0060] This invention involves discharging healthy, initially-stage lithium batteries to different degrees and recording and analyzing the entire aging process of the batteries. This results in more comprehensive data collection and enables the construction of equivalent circuit models for lithium batteries.

[0061] Step 2: Construct an equivalent circuit model of a lithium battery based on the electrochemical principles of the battery, and identify the parameters of the internal equivalent circuit model of the lithium battery.

[0062] Based on the electrochemical principles of lithium-ion batteries, an equivalent circuit model of the battery's internal structure is established. Parameter identification using electrochemical impedance spectroscopy (EIS) determines the parameter values ​​of each component in the equivalent circuit. Considering the movement of Li+ from the negative electrode current collector to the positive electrode current collector in a lithium-ion battery, it needs to pass through the positive and negative electrode double layers, the SEI film, the electrolyte, and the separator. The impedance of Li+ transfer in the current collector, electrolyte, and separator conforms to Ohm's law; therefore, these impedances are often collectively referred to as the battery's ohmic internal resistance, denoted by Re. The impedance of Li+ passing through the SEI film also conforms to Ohm's law, but this part of the impedance is related to the diffusion and migration of Li+ through the insulating layer on the surface of the active material particles. In the EIS, it generally appears as an irregular semicircle; therefore, it can be represented by the SEI resistance R. sei and constant phase angle element Q sei The parallel connection method is used for characterization. Finally, the positive and negative electrode double layers are generated by the aggregation of Li+, and their essence can be represented by the charge transfer internal resistance R. dl With double-layer capacitance C dl The parallel circuit representation is shown, but considering the dispersion effect of EIS in lithium batteries, simply equating the electrode interface double layer to a pure capacitor is not accurate enough. Therefore, it is necessary to introduce the Warburg impedance W for correction. Establishment as follows... Figure 2 Model.

[0063] Among them, the constant phase angle element Q sei The admittance expression is:

[0064] Y Q =Y O (jω) α

[0065] In the formula, Q is a constant phase angle element, Y Q For constant phase angle element admittance, Y O α is the zero-phase admittance, ω is the constant phase angle coefficient, ω is the circuit angular velocity, and j is the imaginary unit.

[0066] The expression for the Warburg impedance is as follows:

[0067]

[0068] In the formula, Z w Let σ be the Warburg impedance value, and σ be the Warburg impedance constant.

[0069] The impedance expression for the equivalent circuit model of a lithium battery is shown below:

[0070]

[0071] In the formula, Z is the impedance of the equivalent circuit model of the lithium battery, and R... e For ohmic internal resistance, R sei For SEI resistor, C dl For double-layer capacitors, Rdl This represents the internal resistance to charge transfer.

[0072] In this embodiment, there are other construction methods for the internal equivalent circuit model of the lithium battery, such as first-order and second-order circuit models.

[0073] The third step involves using the parameters of each component in the equivalent circuit model as unknowns, and fitting the Nyquist plot in the electrochemical impedance spectroscopy of the lithium battery over its entire lifespan using the Levenberg-Marquardt method. This yields the parameters of each component in the equivalent circuit model under different state of health (SOH) of the lithium battery, and then constructs a dataset.

[0074] Electrochemical impedance spectroscopy (EIS) of batteries at different cycle numbers was measured using an electrochemical workstation. With the capacitance and resistance in the model as unknown parameters, the Nyquist plot in the EIS was fitted using the Levenberg-Marquardt method, thereby determining the parameters of each component in the equivalent circuit at different cycle numbers (different state of oxygen).

[0075] This invention establishes a relatively comprehensive internal equivalent circuit model of a lithium battery and uses electrochemical impedance spectroscopy (EIS) for parameter identification. The construction of the lithium battery equivalent circuit model fully considers the structural characteristics of the lithium battery itself, with a focus on modeling and analyzing the SEI film. Using EIS analysis, a small-amplitude sinusoidal potential (current) as the perturbation signal, avoids significant impact on the system. Furthermore, EIS is a frequency domain measurement method, allowing the study of the electrode system using a wide frequency range of impedance spectra, thus providing more kinetic and electrode interface structure information than other conventional methods.

[0076] Step 4: Train the bagged tree algorithm using the dataset and fit the relationship between different health states (SOH), remaining service life (RUL), and the electrochemical impedance spectroscopy of lithium batteries.

[0077] A battery SOH and RUL prediction algorithm was established using machine learning algorithms such as bagged tree modeling. The bagged tree algorithm consists of data sampling, basic model training, and ensemble prediction. The following is a detailed description of each step:

[0078] Data sampling:

[0079] Random sampling: Random sampling with replacement is performed from the original training dataset. This means that each sampling is independent, and each sample may be selected multiple times or not at all. Typically, the number of samples is the same as the size of the original dataset, but due to sampling with replacement, the selected subsamples may contain duplicate samples.

[0080] Creating multiple subsamples: Typically, the bagged tree algorithm creates multiple subsamples, which can be dozens or even hundreds of them. These subsamples are independent, randomly sampled copies.

[0081] For training the base model: Each subsample is used to train a base model, such as a decision tree. Each of these subsamples will be trained with a different set of samples, so each base model will be slightly different.

[0082] Basic model training:

[0083] Independent training: Each subsample is used to train an independent base model. The base model is usually a decision tree, but it can also be other types of classifiers or regression models.

[0084] Model diversity: Because each base model is trained on a different dataset, they will differ from each other, exhibiting a degree of diversity. This helps improve the performance of the ensemble, as different models may capture different patterns in the data.

[0085] Integrated prediction:

[0086] Independent prediction: When a prediction is required, each independent base model will make a prediction on the input data and generate its own prediction results.

[0087] Combined prediction results: For classification problems, a voting method is usually used, that is, selecting the category that receives the most votes as the final prediction result. For regression problems, an averaging method is usually used, that is, averaging the predictions of all basic models as the final prediction result.

[0088] In this implementation, the final integrated prediction result can also be a weighted combination of the prediction results of all basic models, which helps to improve the performance and robustness of the model.

[0089] The bagged tree machine learning algorithm achieves high detection accuracy. The main advantages of the bagged tree algorithm include: reduced variance (by averaging the predictions of multiple base models, bagged trees reduce model variance, improve model stability, and reduce the risk of overfitting); improved generalization ability (since each base model is trained on different subsamples, bagged trees can capture patterns in different subsets of data, thus improving the model's generalization ability); and robustness against noise (by randomly sampling and combining the predictions of multiple models, bagged trees are robust to noise and outliers in the training data).

[0090] Step 5: Substitute the electrochemical impedance spectroscopy of the unknown lithium battery into the trained bag tree algorithm to obtain the state of health (SOH) and remaining lifetime (RUL) of the unknown lithium battery.

[0091] This invention provides a method for predicting State of Health (SOH) and Rullow Limit (RUL) of lithium batteries. Based on the electrochemical principles of batteries, an equivalent circuit model of the battery's internal structure is established. The parameters of each component in the equivalent circuit are treated as unknowns, and the EIS (Electrostatic Interference Sequence) is parameterized. The Levenberg-Marquardt method is used to fit the parameters of components such as ohmic resistance and double-layer capacitance in the equivalent circuit at different cycle numbers. A bagged tree machine learning algorithm is then used to fit the relationship between SOH, RUL, and battery ESI. The bagged tree algorithm, as an ensemble algorithm, is mainly used to randomly sample independent training sets. By averaging the prediction results of multiple basic models, variance can be reduced, model stability improved, and the risk of overfitting reduced. Furthermore, the diversity of basic models can improve the model's generalization ability. Parallel training in each round increases the training speed, making it suitable for fitting the battery's SOH and cycle number in this invention, resulting in more accurate predictions of SOH and RUL.

[0092] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0093] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for predicting the SOH and RUL of a lithium battery, characterized in that, Includes the following steps: Obtain electrochemical impedance spectroscopy data of lithium batteries throughout their entire lifespan; Constructing an equivalent circuit model for lithium batteries based on battery electrochemistry principles; By using the parameters of each component in the equivalent circuit model as unknowns, the Nyquist plot in the full-lifetime electrochemical impedance spectroscopy data of lithium batteries is fitted using the Levenberg-Marquardt method to obtain the parameters of each component in the equivalent circuit model under different health states (SOH) of lithium batteries, and then a dataset is constructed. The bag tree algorithm was trained using a dataset to fit the relationship between different health states (SOH), remaining service life (RUL), and the electrochemical impedance spectroscopy of lithium batteries. By substituting the electrochemical impedance spectroscopy data of the unknown lithium battery into the trained bagged tree algorithm, the state of health (SOH) and remaining lifetime (RUL) of the unknown lithium battery can be obtained.

2. The method for predicting SOH and RUL of a lithium battery as described in claim 1, characterized in that, Electrochemical impedance spectroscopy (EIS) data of lithium batteries over their entire lifespan were obtained through multiple charge-discharge cycles. The steps of one cycle of the charge-discharge experiment include: A 36A constant current discharge of a lithium battery for 5 seconds was used for the inrush current test of ohmic internal resistance. The lithium battery is discharged at a constant current of 20A until its voltage drops below 2.65V; Let the lithium battery stand for 15 minutes to allow the internal chemical reaction to fully occur. Lithium battery 36A constant current charging for 5 seconds; Lithium battery 20A constant current charging for 5 seconds; The lithium battery is charged at a constant current of 20A until its voltage is higher than 3.7V; Let the lithium battery stand for 15 minutes.

3. The method for predicting SOH and RUL of a lithium battery as described in claim 1, characterized in that, The equivalent circuit model inside the lithium battery includes an ohmic internal resistance, an SEI resistor, a constant phase angle element, a charge transfer internal resistance, an electric double layer capacitor, and a Warburg impedance; the SEI resistor is connected in parallel with the constant phase angle element, and the charge transfer internal resistance is connected in series with the Warburg impedance and then in parallel with the electric double layer capacitor.

4. The method for predicting SOH and RUL of a lithium battery as described in claim 3, characterized in that, The admittance expression for the constant phase angle element is as follows: Y Q = Y O (jω) α In the formula, Q is a constant phase angle element, Y Q For constant phase angle element admittance, Y O α is the zero-phase admittance, ω is the constant phase angle coefficient, ω is the circuit angular velocity, and j is the imaginary unit. The Warburg impedance expression is as follows: In the formula, Z w Let σ be the Warburg impedance value, and σ be the Warburg impedance constant.

5. The method for predicting SOH and RUL of a lithium battery as described in claim 4, characterized in that, The impedance expression of the equivalent circuit model of the lithium battery is shown below: In the formula, Z is the impedance of the equivalent circuit model of the lithium battery, and R... e For ohmic internal resistance, R sei For SEI resistor, C dl For double-layer capacitors, R dl This represents the internal resistance to charge transfer.

6. The method for predicting SOH and RUL of a lithium battery as described in claim 1, characterized in that, Training the bagged tree algorithm using a dataset includes the following steps: Data sampling is performed on the dataset; Multiple subsamples are obtained by random sampling with replacement from the dataset. Each of the sub-samples is used to train a base model; Multiple basic models make predictions on the input data, generating multiple prediction results; The final prediction result is obtained by combining multiple prediction results.

7. The method for predicting SOH and RUL of a lithium battery as described in claim 6, characterized in that, The basic model is a decision tree, classifier, or regression model.

8. The method for predicting SOH and RUL of a lithium battery as described in claim 6, characterized in that, Multiple prediction results are combined to obtain the final prediction result, including: For classification problems, the category that receives the most votes is selected as the prediction result; For regression problems, the average of the predictions from all basic models is taken as the prediction result.

Images