A battery state of health estimation method based on electrochemical impedance spectroscopy
By measuring the electrochemical impedance spectroscopy curve of the battery and constructing a machine learning model, the problem of reduced parameter accuracy of the equivalent circuit model during battery aging was solved, and accurate estimation of battery health status was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI AEROSPACE POWER TECH
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-28
AI Technical Summary
In the existing technology, the accuracy of parameters in the equivalent circuit model decreases during battery aging, resulting in a decrease in the accuracy of battery health state estimation and making it difficult to accurately describe the battery aging process.
By measuring the electrochemical impedance spectroscopy curves of batteries under different states of charge, feature points are extracted, a machine learning model is constructed, and the battery health status is estimated based on the electrochemical impedance spectroscopy. The random forest algorithm is used to train and test the model.
It improves the accuracy and precision of battery health status estimation, reduces estimation errors, and provides a more reliable data basis for battery aging characteristics.
Smart Images
Figure CN122469205A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of battery health state estimation technology, and specifically to a battery health state estimation method. Background Technology
[0002] Against the backdrop of actively promoting carbon neutrality, batteries (including lithium-ion batteries, sodium-ion batteries, quasi-solid-state batteries, and all-solid-state batteries) are widely used in numerous fields as key energy supply components. Whether in electric vehicles, renewable energy systems, or various portable electronic devices, the performance and state of health (SOH) of the battery directly affect the safety and economy of the entire system; therefore, accurately understanding the battery's state of health is crucial. Some existing research methods simplify the complex electrochemical processes within batteries, estimating the battery's state of health by constructing equivalent circuit models. However, as the number of charge-discharge cycles increases, the accuracy of the equivalent circuit model parameters decreases significantly, making it difficult to accurately describe battery aging and thus reducing the accuracy of battery state of health estimation.
[0003] In a laboratory setting, by applying small-amplitude AC signals of different frequencies to a battery, the ratio of AC voltage to current (i.e., the system impedance) is measured as a function of sinusoidal frequency. Changes in these changes can lead to the acquisition of the battery's electrochemical impedance spectroscopy (EIS). EIS reflects the battery's characteristics across different frequency ranges, and these frequency characteristics are closely related to battery aging. Therefore, using battery EIS to estimate state of health (SOH) is gradually becoming an important direction in battery health status research. Summary of the Invention
[0004] The purpose of this invention is to provide a battery health state estimation method based on electrochemical impedance spectroscopy, thereby solving the above-mentioned technical problems; The technical problem solved by this invention can be achieved by the following technical solutions: A method for estimating the health status of a battery based on electrochemical impedance spectroscopy includes, Step S1: Take the average of the charging capacity and discharging capacity of the battery after a certain number of charge-discharge cycles to obtain the capacity tag of the current cycle, and calculate the health status tag of the current cycle based on the capacity tag of the current cycle. Step S2: Measure the electrochemical impedance spectroscopy curves of the battery at several states of charge under corresponding cycles, and extract the characteristic points on each electrochemical impedance spectroscopy curve. Step S3: Arrange the health status labels and feature points corresponding to the N iterations to obtain a dataset sequence; Step S4: A battery health status estimation model is constructed using a machine learning algorithm. The battery health status estimation model is trained and tested based on the dataset sequence. The trained battery health status estimation model is then used to estimate the battery health status. Where N is a positive integer.
[0005] Preferably, in step S2, electrochemical impedance spectroscopy data are measured multiple times under each charged state, and the average value of the multiple measured electrochemical impedance spectroscopy data is taken to obtain the average electrochemical impedance spectroscopy curve under each charged state, and the average electrochemical impedance spectroscopy curve is used as the electrochemical impedance spectroscopy curve.
[0006] Preferably, step S2 includes extracting the feature points of the electrochemical impedance spectroscopy curve of the battery in the first state of charge to obtain the first feature point. The extraction method is to determine whether there is an intersection between the electrochemical impedance spectroscopy curve and the real axis. If there is an intersection, the intersection between the electrochemical impedance spectroscopy curve and the real axis is extracted as the first feature point. If there is no intersection, the point with the smallest absolute value of the imaginary part in the electrochemical impedance spectroscopy curve is selected as the first feature point.
[0007] Preferably, step S2 includes extracting the feature points of the electrochemical impedance spectroscopy curve of the battery in the second state of charge to obtain the second feature point. The extraction method is to determine whether there is an intersection between the electrochemical impedance spectroscopy curve and the real axis. If there is an intersection, the intersection between the electrochemical impedance spectroscopy curve and the real axis is extracted as the second feature point. If there is no intersection, the point with the smallest absolute value of the imaginary part in the electrochemical impedance spectroscopy curve is selected as the second feature point.
[0008] Preferably, in step S3, the expression for the dataset sequence is:
[0009] in Indicates the first The next loop Indicates the first The health status label for the next cycle. Indicates the first The first feature point in the next cycle, Indicates the first The second feature point in the next cycle.
[0010] Preferably, in step S1, the calculation formula for the health status tag in the current cycle is:
[0011] in The health status label represents the current cycle. The capacity label represents the current cycle. The capacity label indicates the initial loop.
[0012] Preferably, step S1 further includes, when the battery loses charging or discharging capacity after a cycle, retaining the recorded charging or discharging capacity as the capacity tag for the current cycle.
[0013] Preferably, in step S4, two battery health state estimation models with the same parameters are trained separately to obtain two trained battery health state estimation models that estimate the battery health state respectively. The training method includes... The first battery health status estimation model is trained and tested using the dataset sequence of several cycles of a first number of batteries as the training set and the health status labels and feature points of several cycles of a second number of batteries as the test set. The second battery health status estimation model is trained and tested using the dataset sequence of several cycles of a third number of batteries as the training set and the health status labels and feature points of several cycles of a fourth number of batteries as the test set.
[0014] The beneficial effects of the present invention are as follows: By adopting the above technical solution, the present invention obtains the electrochemical impedance spectrum of the battery, and then constructs a machine learning model for training. Based on the characterization of the electrochemical impedance spectrum in the battery aging process, the health status of the battery can be accurately estimated. Attached Figure Description
[0015] Figure 1 This is a schematic diagram illustrating the steps of the battery health state estimation method based on electrochemical impedance spectroscopy in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the extraction of the first feature point when the battery is in a first state of charge, as shown in an embodiment of the present invention. Figure 3 This is a schematic diagram illustrating the extraction of the second feature point when the battery is in a second state of charge, according to an embodiment of the present invention. Detailed Implementation
[0016] 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.
[0017] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.
[0018] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.
[0019] A method for estimating battery health status based on electrochemical impedance spectroscopy, such as Figure 1 As shown, including, Step S1: Take the average of the charging capacity and discharging capacity of the battery after a certain number of charge-discharge cycles to obtain the capacity tag of the current cycle, and calculate the health status tag of the current cycle based on the capacity tag of the current cycle. Step S2: Measure the electrochemical impedance spectroscopy curves of the battery at several states of charge under corresponding cycles, and extract the characteristic points on each electrochemical impedance spectroscopy curve. Step S3: Arrange the health status labels and feature points of N cycles to obtain the dataset sequence; Step S4: A battery health status estimation model is constructed using a machine learning algorithm. The battery health status estimation model is trained and tested based on the dataset sequence. The trained battery health status estimation model is then used to estimate the battery health status. Where N is a positive integer.
[0020] Specifically, in laboratory measurements, a battery can be equivalently represented as a circuit model composed of resistors, capacitors, and inductors. At different stages of the battery's cycle life, a small-amplitude AC signal of different frequencies is applied to the battery. The ratio of AC voltage to current (this ratio represents the system impedance) is measured as a function of the sinusoidal frequency ω to obtain the battery's electrochemical impedance spectroscopy (EIS), thereby estimating the battery's state of health. Since the EIS characterizes the battery's state of health across different frequency ranges during aging, reflecting its condition, this invention proposes a battery health estimation method based on EIS. In a preferred embodiment, in step S2, electrochemical impedance spectroscopy data are measured multiple times under each charged state, and the average value of the multiple measured electrochemical impedance spectroscopy data is taken to obtain the average electrochemical impedance spectroscopy curve under each charged state, and the average electrochemical impedance spectroscopy curve is used as the electrochemical impedance spectroscopy curve.
[0021] Specifically, taking multiple measurements of electrochemical impedance spectroscopy data and averaging them can effectively reduce the influence of random factors during the measurement process, making the obtained average electrochemical impedance spectroscopy curve more stable and reliable.
[0022] In actual measurements, the state of a battery is affected by various random factors such as ambient temperature and minor fluctuations in the measuring instrument, resulting in variations in measurement results each time. By averaging these random errors, the final average curve more accurately reflects the battery's true electrochemical characteristics, providing a more reliable data basis for subsequent health state estimation.
[0023] When constructing the dataset, more accurate electrochemical impedance spectroscopy data allows for more precise feature point extraction, thereby improving the accuracy of the health status estimation model trained on this data and reducing estimation errors. In subsequent steps, feature points are extracted based on the intersection of the average curve with the real axis or the point with the smallest absolute value of the imaginary part. The average curve can more clearly and stably present these features, avoiding errors in feature point selection due to fluctuations in a single measurement curve, thus improving the accuracy of feature extraction and laying the foundation for building a more accurate health status estimation model.
[0024] More specifically, during the testing process, this invention selected two extreme states of charge (0% and 100%) for measurement. State of charge refers to the percentage of remaining charge in the battery relative to its rated capacity; 0% indicates a fully discharged battery, and 100% indicates a fully charged battery.
[0025] The present invention selects the states of charge as 0% and 100% because these two extreme states can reflect the electrochemical characteristics of the battery under extreme conditions, thereby obtaining a more comprehensive understanding of the battery's performance.
[0026] This invention involves measuring electrochemical impedance spectroscopy data three times at each state of charge (SOC). The purpose of multiple measurements is to reduce measurement error.
[0027] By averaging the EIS data obtained from three measurements, the average EIS curves for the corresponding cycles at 0% and 100% state of charge can be obtained. The resulting average curve is used to represent the electrochemical impedance characteristics of the battery at that state of charge.
[0028] In a preferred embodiment, step S2 includes extracting feature points from the electrochemical impedance spectroscopy curve of the battery when it is in the first state of charge of 0% to obtain the first feature point. The extraction method is to determine whether there is an intersection between the electrochemical impedance spectroscopy curve and the real axis. If there is an intersection, the intersection between the electrochemical impedance spectroscopy curve and the real axis is extracted as the first feature point. If there is no intersection, the point with the smallest absolute value of the imaginary part in the electrochemical impedance spectroscopy curve is selected as the first feature point.
[0029] Specifically, such as Figure 2 As shown, based on the EIS average curve with a state of charge of 0%, feature points are extracted using the obtained EIS average curve with a state of charge of 0%.
[0030] Intersection Priority Principle: First, try to extract the intersection point of the curve with the real axis as the first feature point. In the EIS spectrum, the real axis represents resistance, and the intersection point of the curve with the real axis reflects the DC resistance characteristic of the battery under that state of charge. Replacement by the point with the minimum absolute value of the imaginary part: The imaginary part reflects the capacitance and inductance characteristics inside the battery. If the curve does not intersect the real axis, the point with the minimum absolute value of the imaginary part is selected as the first characteristic point, denoted as the first characteristic point. .
[0031] In a preferred embodiment, step S2 includes extracting feature points from the electrochemical impedance spectroscopy curve of the battery when it is in the second state of charge of 100% to obtain the second feature point. The extraction method is to determine whether there is an intersection between the electrochemical impedance spectroscopy curve and the real axis. If there is an intersection, the intersection between the electrochemical impedance spectroscopy curve and the real axis is extracted as the second feature point. If there is no intersection, the point with the smallest absolute value of the imaginary part in the electrochemical impedance spectroscopy curve is selected as the second feature point.
[0032] Similarly, feature points are extracted based on the average EIS curve when the state of charge is 100%.
[0033] Intersection priority principle: Prioritize extracting the intersection point of the curve with the real axis, which reflects the DC resistance characteristics of the battery when it is fully charged.
[0034] Replacement by the point with the minimum absolute value of the imaginary part: If the curve has no intersection with the real axis, then select the point with the minimum absolute value of the imaginary part of the curve as the second characteristic point, denoted as . .
[0035] In a preferred embodiment, in step S3, the expression for the dataset sequence is:
[0036] in Indicates the first The next loop Indicates the first The health status label for the next cycle. Indicates the first The first feature point of the next cycle Indicates the first The second feature point of the next cycle.
[0037] Specifically, in the process of estimating battery health status based on electrochemical impedance spectroscopy (EIS), a large amount of data is needed to train and validate the model to establish the relationship between EIS features and battery health status. This invention constructs a dataset sequence to integrate the battery health status under different cycles and the corresponding EIS feature points, providing input data for the machine learning model, thereby achieving accurate estimation of battery state of health (SOH).
[0038] In a preferred embodiment, in step S1, the formula for calculating the health status label of the current cycle is:
[0039] in The label represents the health status of the current cycle. The capacity label indicates the current loop. The capacity label indicates the initial loop.
[0040] In a preferred embodiment, step S1 further includes retaining the recorded charging or discharging capacity as the capacity tag for the current cycle when the battery is missing charging or discharging capacity after a cycle.
[0041] Specifically, after a certain number of charge-discharge cycles, the battery's charging and discharging capacities show a significant decrease. The average of the charging and discharging capacities is used to obtain the battery's capacity label for the current cycle. Since some batteries lack charging / discharging capacity for certain cycles, the recorded charging / discharging capacity is retained as the capacity label for the current cycle for these missing values. Based on the current capacity label, the SOH (State of Health) label for the current cycle can be calculated. The formula for calculating the health status label in the current cycle is:
[0042] in The label represents the health status of the current cycle. The capacity label indicates the current loop. The capacity label indicates the initial loop.
[0043] In a preferred embodiment, in step S4, a battery health status estimation model is constructed using the Random Forest (RF) algorithm.
[0044] In a preferred embodiment, in step S4, two battery health state estimation models with identical parameters are trained separately to obtain two trained battery health state estimation models that estimate the battery health state respectively. The training method includes... The first battery health status estimation model is trained and tested using a dataset sequence of several cycles of the first number of batteries as the training set and the health status labels and feature points of several cycles of the second number of batteries as the test set. The second battery health status estimation model was trained and tested using a dataset sequence of several cycles from a third number of batteries as the training set and health status labels and feature points from several cycles from a fourth number of batteries as the test set.
[0045] Specifically, the battery health state estimation model employs the random forest algorithm. Based on the existing average curve of electrochemical impedance spectroscopy, and with the algorithm model parameters remaining unchanged, two sets of model training methods are set up, and then battery health state estimation is performed separately for each set.
[0046] In this embodiment, the training methods for the two sets of models are as follows: Training 1 uses the dataset sequences of the 6th, 11th, 16th, 21st, and 26th cycles of the 3 batteries as the training set, and the SOH labels and EIS feature input sequences of the 6th, 11th, 16th, 21st, and 26th cycles of the 2 batteries as the test set.
[0047] Training 2 uses the dataset sequences of the 6th, 11th, 16th, 21st, and 26th cycles of 4 batteries as the training set, and the SOH labels and EIS feature input sequences of the 6th, 11th, 16th, 21st, and 26th cycles of 1 battery as the test set.
[0048] Based on the training set, the battery health state estimation model learns the relationship between the EIS feature input sequence and the SOH label, adjusts its parameters, and constructs a mapping relationship. Based on the test set, the EIS feature input sequence is input into the trained battery health state estimation model, which predicts the corresponding SOH value. The performance of the battery health state estimation model is then evaluated by comparing the predicted value with the actual SOH label.
[0049] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for estimating the health status of a battery based on electrochemical impedance spectroscopy, characterized in that, include, Step S1: Take the average of the charging capacity and discharging capacity of the battery after a certain number of charge-discharge cycles to obtain the capacity tag of the current cycle, and calculate the health status tag of the current cycle based on the capacity tag of the current cycle. Step S2: Measure the electrochemical impedance spectroscopy curves of the battery at several states of charge under corresponding cycles, and extract the characteristic points on each electrochemical impedance spectroscopy curve. Step S3: Arrange the health status labels and feature points corresponding to the N iterations to obtain a dataset sequence; Step S4: A battery health status estimation model is constructed using a machine learning algorithm. The battery health status estimation model is trained and tested based on the dataset sequence. The trained battery health status estimation model is then used to estimate the battery health status. Where N is a positive integer.
2. The battery health state estimation method based on electrochemical impedance spectroscopy according to claim 1, characterized in that, In step S2, electrochemical impedance spectroscopy data are measured multiple times under each charged state. The average value of the electrochemical impedance spectroscopy data obtained from the multiple measurements is taken to obtain the average electrochemical impedance spectroscopy curve under each charged state, and the average electrochemical impedance spectroscopy curve is used as the electrochemical impedance spectroscopy curve.
3. The battery health state estimation method based on electrochemical impedance spectroscopy according to claim 1, characterized in that, Step S2 includes extracting the feature points of the electrochemical impedance spectroscopy curve of the battery in the first state of charge to obtain the first feature point. The extraction method is to determine whether there is an intersection between the electrochemical impedance spectroscopy curve and the real axis. If there is an intersection, the intersection between the electrochemical impedance spectroscopy curve and the real axis is extracted as the first feature point. If no intersection point exists, the point with the smallest absolute value of the imaginary part in the electrochemical impedance spectroscopy curve is selected as the first feature point.
4. The battery health state estimation method based on electrochemical impedance spectroscopy according to claim 1, characterized in that, Step S2 includes extracting the feature points of the electrochemical impedance spectroscopy curve of the battery in the second state of charge to obtain the second feature points. The extraction method is to determine whether there is an intersection between the electrochemical impedance spectroscopy curve and the real axis. If there is an intersection, the intersection between the electrochemical impedance spectroscopy curve and the real axis is extracted as the second feature points. If no intersection point exists, the point with the smallest absolute value of the imaginary part in the electrochemical impedance spectroscopy curve is selected as the second feature point.
5. The battery health state estimation method based on electrochemical impedance spectroscopy according to claim 4, characterized in that, In step S3, the expression for the dataset sequence is: ； in Indicates the first The next loop. Indicates the first The health status label for the next cycle. Indicates the first The first feature point in the next cycle, Indicates the first The second feature point in the next cycle.
6. The battery health state estimation method based on electrochemical impedance spectroscopy according to claim 1, characterized in that, In step S1, the formula for calculating the health status label in the current loop is: ； in The health status label represents the current cycle. The capacity label represents the current cycle. The capacity label indicates the initial loop.
7. The battery health state estimation method based on electrochemical impedance spectroscopy according to claim 1, characterized in that, Step S1 also includes that when the battery loses charging or discharging capacity after a cycle, the recorded charging or discharging capacity is retained as the capacity tag for the current cycle.
8. The battery health state estimation method based on electrochemical impedance spectroscopy according to claim 1, characterized in that, In step S4, two battery health state estimation models with identical parameters are trained separately to obtain two trained battery health state estimation models that estimate the battery health state. The training method includes... The first battery health status estimation model is trained and tested using the dataset sequence of several cycles of a first number of batteries as the training set and the health status labels and feature points of several cycles of a second number of batteries as the test set. The second battery health status estimation model is trained and tested using the dataset sequence of several cycles of a third number of batteries as the training set and the health status labels and feature points of several cycles of a fourth number of batteries as the test set.