A method for diagnosing the health status of lithium-ion batteries based on equivalent circuit feature screening.
By combining equivalent circuit models with data-driven features to estimate the state of health of lithium-ion batteries, the problems of low prediction accuracy and high computational complexity in existing technologies are solved, and efficient and interpretable battery health management is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2025-03-13
- Publication Date
- 2026-05-26
AI Technical Summary
Existing methods for predicting the health status of lithium-ion batteries suffer from limitations in prediction accuracy, high computational complexity, insufficient generalization ability, and a lack of physical interpretability in data-driven methods, resulting in low accuracy and efficiency of battery management systems.
By combining the features of the equivalent circuit model with data-driven features, redundant information is reduced through feature selection. A machine learning model is used to estimate the health status of lithium-ion batteries. Physical modeling is introduced to improve prediction accuracy and computational efficiency, and to enhance the generalization ability of the model.
It improves the accuracy and computational efficiency of lithium-ion battery health status estimation, enhances the interpretability of the model, and can adapt to different operating conditions and environmental changes, meeting the needs of real-time or online prediction.
Smart Images
Figure CN120064998B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of energy storage battery health status estimation, and more specifically, relates to a method for estimating the health status of energy storage batteries based on equivalent circuit feature screening and utilizing machine learning algorithms and data models. Background Technology
[0002] With the increasing prevalence of lithium-ion batteries (LIBs) in modern society, the safety risks and range anxiety they cause are becoming increasingly prominent. The performance degradation and aging process of LIBs can easily lead to system failure, posing risks of property damage and personal injury. Range anxiety often stems from the mismatch between the mileage displayed on the dashboard and the actual remaining battery capacity. Building an accurate battery health status diagnostic system has become a key path to solving these technical bottlenecks. Battery state of health (SOH) prediction has received widespread attention. Currently, mainstream SOH scalar characterization methods are mainly based on historical cycle discharge capacity or internal resistance parameters. Specifically, in energy supply-centric applications such as electric vehicles, capacity parameters are often used as SOH indicators; while internal resistance parameters are mostly used in applications such as hybrid vehicles where power output is the core requirement. To achieve these goals, existing technologies have proposed three types of solutions: model-driven, data-driven, and hybrid methods. However, current lithium-ion battery SOH estimation faces problems such as limited prediction accuracy, high computational complexity, insufficient generalization ability, and a lack of physical interpretability in data-driven methods. The presence of redundant features may affect model performance and lead to low computational efficiency. While pure data-driven methods can provide some predictive ability, they are difficult to reveal the physical degradation mechanism of battery health. Summary of the Invention
[0003] To address these issues, this method combines equivalent circuit model (ECM) features with data-driven features. By reducing redundant information through feature selection, it improves the accuracy and computational efficiency of SOH estimation, while enhancing the model's generalization ability to adapt to different operating conditions and environmental changes. Compared to traditional black-box models, this method introduces physical modeling, making the prediction results more closely reflect the actual degradation process of batteries, improving interpretability. Furthermore, by optimizing feature inputs, it reduces computational complexity, enabling it to meet real-time or online prediction requirements. By integrating the advantages of data-driven and physical modeling, this method achieves optimization in reliability, accuracy, and applicability, providing an efficient and feasible solution for lithium-ion battery health management.
[0004] This invention discloses a method for diagnosing the health status of lithium-ion batteries based on equivalent circuit feature screening, comprising the following steps:
[0005] Step 1: Select a battery dataset with high data quality, sufficient sample size, and strong representativeness;
[0006] Step 2: Preprocess the data and extract features;
[0007] Step 3: Input the features into the machine learning model to perform the first assessment of the battery health status;
[0008] Step 4: Establish an equivalent circuit model and extract relevant parameters, couple the parameters with incremental capacity curve (ICA) features, and further filter features;
[0009] Step 5: Input the filtered feature pairs to reassess the battery health status.
[0010] In the above technical solution, step 1 specifically includes:
[0011] The battery dataset used in this invention is from the Zenodo database, which includes charge / discharge data and frequency domain EIS (electrochemical impedance spectroscopy) data. The study selected a graphite / NCA 18650 battery (Dataset_1) as the object of study, with a nominal capacity of 3.5 Ah, a rated voltage of 3.6 V, a charging cutoff voltage of 4.2 V, and a discharging cutoff voltage of 2.65 V.
[0012] In the above technical solution, step 2 specifically includes:
[0013] Step 2.1: The target value of this invention is the battery's health status data, which can be represented as:
[0014]
[0015] Q n Q0 represents the capacity of the battery when it is fully charged (State of Charge = 100%), and Q0 represents the rated capacity of the battery.
[0016] Step 2.2: The main method for handling outliers is to use the Z-score method to filter out outliers. Data that exceeds the threshold is replaced by the average of the previous and subsequent values.
[0017] Step 2.3: Extract features from the charging data, including: constant current charging time, constant voltage charging time, initial voltage, average voltage, constant current charging time / constant voltage charging time, and constant current charging time / total charging time, thus obtaining 6 different features;
[0018] Step 2.4: Process the charging data to obtain the incremental capacity curve;
[0019] Step 2.5: If the peak disappears in the curve, find the point with the smallest absolute value of the derivative in the voltage window and take it as the value of the peak.
[0020] Step 2.6: Select the peak value and peak voltage of each peak (there are three peaks in the curve) of the incremental capacity curve after processing the peak disappearance as the other 6 features input to the machine learning model.
[0021] In the above technical solution, step 3 specifically includes:
[0022] Step 3.1: Input the 12 features extracted above into the machine learning model. The machine learning models include: LightGBM model, XGBoost model, GPR model and Ridge model.
[0023] Step 3.2: Under different temperature conditions, set up different training and testing paths, and adopt a cross-validation strategy: at 25°C, use four batteries: 25C03, 25C06, 25C11, and 25C19, each as the test set, and the remaining batteries as the training set; and so on, at 35°C and 45°C, follow the same path, but with different batteries. At 35°C, use batteries 35C01 and 35C02; at 45°C, use batteries 45C13, 45C16, 45C21, and 45C27 respectively. This cross-validation framework effectively ensures the model's generalization ability to different individual batteries of the same model; the input is the 12 features obtained above, and the output is the battery health status;
[0024] Step 3.3: Use commonly used machine learning evaluation metrics, including: mean squared error (MSE), root mean squared error (RMSE), and coefficient of determination (R²).
[0025] In the above technical solution, step 4 specifically includes:
[0026] Step 4.1: Extract the battery impedance data: including 100 data points in the frequency range of 50 mHz to 10 kHz, and extract the data once every 25 cycles (impedance test cycle);
[0027] Step 4.2: Based on the obtained impedance data, establish and fit an equivalent circuit model using the `impedance.py` module of the Python library. The equivalent circuit model follows... Figure 3 Perform modeling;
[0028] Step 4.3: Extract the parameters most relevant to battery degradation from the equivalent circuit model: R sei C sei C ct R ct Z w ;
[0029] Step 4.4: Perform Pearson correlation analysis on the 12 features extracted in Step 2 and these 5 parameters. Sum the Pearson coefficients of each feature and the parameters, take the average, and sort them. Remove the features that are ranked lower or have an average value less than 0.6 to obtain the filtered features.
[0030] In the above technical solution, step 5 specifically includes:
[0031] Step 5.1: Similar to the scheme in Step 3 above, a cross-validation strategy is adopted, with the input being the filtered features and the output being the battery health status;
[0032] Step 5.2: Evaluate the results using the same machine learning evaluation metrics as in Step 3 above.
[0033] Because the present invention employs the above-mentioned technical means, it has the following beneficial effects:
[0034] This method combines equivalent circuit model features with data-driven features, utilizing feature selection to reduce redundant information, thereby improving the accuracy and computational efficiency of SOH estimation. It also enhances the model's generalization ability, enabling it to adapt to different operating conditions and environmental changes. Compared to purely data-driven methods, this method introduces ECM features, making the prediction results more closely reflect the battery's physical degradation mechanism and improving the model's interpretability. Furthermore, optimized feature selection reduces computational complexity, making SOH estimation more efficient and meeting real-time or online prediction requirements. By integrating the advantages of data-driven and physical modeling, this method combines reliability and accuracy, providing a superior solution for lithium-ion battery health management. Attached Figure Description
[0035] Figure 1 A flowchart illustrating the implementation steps of the estimation method provided in this embodiment of the invention;
[0036] Figure 2 Battery degradation curves of the dataset provided in this embodiment of the invention;
[0037] Figure 3 A schematic diagram illustrating the establishment of an equivalent circuit model provided in an embodiment of the present invention;
[0038] Figure 4 A comparison chart of the estimated and actual SOH values of a 25C11 battery at 25°C, provided for embodiments of the present invention. Detailed Implementation
[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0040] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0041] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. It should be noted that, unless otherwise specified, features in the embodiments of the present invention can be combined with each other.
[0042] This invention discloses a method for diagnosing the health status of lithium-ion batteries based on equivalent circuit feature screening (hereinafter referred to as: estimation method). The estimation method includes the following steps:
[0043] Step 1: Select a battery dataset with high data quality, sufficient sample size, and strong representativeness;
[0044] Step 2: Preprocess the data and extract features;
[0045] Step 3: Input the features into the machine learning model to perform the first assessment of the battery health status;
[0046] Step 4: Establish an equivalent circuit model and extract relevant parameters, couple the parameters with ICA features, and further filter the features;
[0047] Step 5: Input the filtered feature pairs to reassess the battery health status.
[0048] Example 1
[0049] A method for diagnosing the health status of lithium-ion batteries based on equivalent circuit feature screening includes the following steps:
[0050] Step 1: Select a battery dataset with high data quality, sufficient sample size, and strong representativeness.
[0051] In the above technical solution, step 1 specifically includes:
[0052] The battery dataset used in this invention is sourced from the Zenodo database, containing charge / discharge data and frequency-domain EIS (electrochemical impedance spectroscopy) data. The study selected a graphite / NCA 18650 battery (Dataset_1) as the object, with a nominal capacity of 3.5 Ah, a rated voltage of 3.6 V, a charging cut-off voltage of 4.2 V, and a discharging cut-off voltage of 2.65 V. The battery's charge / discharge protocol covers three rates: 0.5C, 1C, and 2C. The cycling temperature is set to a wide range of -20°C to 25°C. The specific test scheme includes capacity calibration cycling, dynamic operating condition cycling, and EIS testing at different cycle intervals. The lithium battery charging protocol adopts a standard constant current-constant voltage (CC-CV) strategy. The charging process is divided into three parts: (I) constant current charging stage, (II) constant voltage charging stage, and (III) post-charging resting stage. The specific process is as follows: First, the battery is charged at a constant current rate from 0.25C (0.875 A) to 1C (3.5 A) to 4.2 V. Then, it is switched to 4.2 V constant voltage charging until the current drops to 0.05C (0.175 A), with a sampling interval of 10 s. After charging, the battery is left to stand for 30 minutes, with a sampling interval of 2 minutes. The discharge protocol adopts a 1C constant current discharge mode, and the discharge is terminated when the voltage drops to 2.65 V. The cycle temperature (25°C, 35°C, 45°C) is controlled by a high-precision constant temperature chamber (±0.2°C). For impedance data, the lithium-ion battery electrochemical impedance spectroscopy (EIS) test frequency range is 50 mHz to 10 kHz (20 data points are collected every ten octaves), and the excitation signal potential amplitude is 20 mV. The EIS test is performed under steady-state conditions after a 30-minute rest following full charge, and is executed at intervals of 25 charge-discharge cycles. Figure 2 As shown, the capacity of all 10 commercial lithium-ion batteries used in this study showed a significant decreasing trend with increasing cycle number.
[0053] Step 2: Preprocess the data and extract features.
[0054] In the above technical solution, step 2 specifically includes:
[0055] Step 2.1: The target value of this invention is the battery's health status data, which can be represented as:
[0056]
[0057] Qn is the capacity of the battery when it is fully charged (SOC = 100%), and Q0 is the rated capacity of the battery.
[0058] Step 2.2: The main method for handling outliers is to use the Z-score method to filter out outliers. Data that exceeds the threshold is replaced by the average of the previous and subsequent values.
[0059] Step 2.3: Extract features from the charging data, including: constant current charging time, constant voltage charging time, initial voltage, average voltage, constant current charging time / constant voltage charging time, and constant current charging time / total charging time, thus obtaining 6 different features;
[0060] Step 2.4: Process charging data to obtain the incremental capacity curve; the incremental capacity analysis method transforms the voltage plateau of traditional constant current charging into easily observable peaks and valleys on the incremental capacity curve to correspond to the aging state and aging mechanism inside the battery. The specific calculation method is as follows: the incremental capacity is equal to the ratio of the difference in charge input to the difference in voltage between two consecutive moments:
[0061]
[0062] Among them, Q k To charge the device at time k, Q k-1 Charge the energy at time k-1. V k Let V be the voltage value at time k. k-1 This represents the voltage value at time k-1.
[0063] Step 2.5: If the peak disappears in the curve, find the point with the smallest absolute value of the derivative in the voltage window and take it as the value of the peak.
[0064] Step 2.6: Select the peak value and peak voltage of each peak (there are three peaks in the curve) of the incremental capacity curve after processing the peak disappearance as the other 6 features input to the machine learning model.
[0065] Step 3: Input the features into the machine learning model to perform the first assessment of the battery health status.
[0066] In the above technical solution, step 3 specifically includes:
[0067] Step 3.1: Input the 12 features extracted above into the machine learning models, including: LightGBM model, XGBoost model, GPR model, and Ridge model; the principles of these models are as follows:
[0068] Ridge model:
[0069] Ridge regression is a regularization method for linear regression designed to address the overfitting problem of ordinary least squares (OLS) in multicollinearity or high-dimensional data. Its core principle is to apply L2 regularization constraints to the model's coefficients, thereby reducing model complexity and improving generalization ability.
[0070] XGBoost (Extreme gradient boosting) model:
[0071] XGBoost is an optimized implementation based on Gradient Boosting Decision Trees (GBDT). It employs the Boosting ensemble learning framework to improve the model's predictive ability by weighted combination of multiple decision trees. Its basic idea is to train one tree, calculate the residual (i.e., the error between the predicted and actual values), train the next tree to fit the residual of the previous tree, and iterate in this way until a preset number of trees is reached or the error converges.
[0072] LightGBM (Light gradient boosting machine) model:
[0073] LightGBM, also a variant of GBDT, is specifically designed to address the limitations of GBDT when handling massive datasets, effectively improving its practical application value and computational efficiency. LightGBM's decision tree model employs a leaf-growth splitting strategy, which has relatively low computational complexity but requires limiting tree depth and the minimum sample size of leaf nodes to suppress overfitting. Furthermore, LightGBM uses a histogram-based decision tree algorithm, discretizing feature values into several "buckets." By searching for split points within these "buckets," storage requirements and computational consumption are significantly reduced. The algorithm's parallelization framework includes three optimization strategies: feature parallelism when feature dimensions are high, data parallelism when the data scale is large, and voting parallelism when both feature dimensions and the number of votes are high.
[0074] Step 3.2: Under different temperature conditions, set up different training and testing paths, and adopt a cross-validation strategy: at 25°C, use four batteries: 25C03, 25C06, 25C11, and 25C19, each as the test set, and the remaining batteries as the training set; and so on, at 35°C and 45°C, follow the same path, but with different batteries. At 35°C, use batteries 35C01 and 35C02; at 45°C, use batteries 45C13, 45C16, 45C21, and 45C27 respectively. This cross-validation framework effectively ensures the model's generalization ability to different individual batteries of the same model; the input is the 12 features obtained above, and the output is the battery health status;
[0075] Step 3.3: Use commonly used machine learning evaluation metrics, including: mean squared error (MSE), root mean squared error (RMSE), and coefficient of determination (R²).
[0076] Step 4: Establish an equivalent circuit model and extract relevant parameters, couple the parameters with ICA features, and further filter the features;
[0077] In the above technical solution, step 4 specifically includes:
[0078] Step 4.1: Extract the battery impedance data: including 100 data points in the frequency range of 50 mHz to 10 kHz, and extract the data once every 25 cycles (impedance test cycle);
[0079] Step 4.2: Based on the obtained impedance data, establish and fit an equivalent circuit model using the `impedance.py` module in Python. The equivalent circuit model follows... Figure 3 Perform modeling;
[0080] Step 4.3: Extract the parameters most relevant to battery degradation from the equivalent circuit model: R sei C sei C ct R ct Z w R sei / / C sei and R ct / / C ct Represents the loss of lithium inventory (LAM); impedance Z w Used to describe the loss of active material (LAM).
[0081] Step 4.4: Perform Pearson correlation analysis on the 12 features extracted in Step 2 and these 5 parameters. Sum the Pearson coefficients of each feature and the parameters, take the average, and sort them. Remove the features that are ranked lower or have an average value less than 0.6 to obtain the filtered features.
[0082] Step 5: Input the filtered feature pairs to reassess the battery health status.
[0083] In the above technical solution, step 5 specifically includes:
[0084] Step 5.1: Similar to the scheme in Step 3 above, a cross-validation strategy is adopted, with the input being the filtered features and the output being the battery health status;
[0085] Step 5.2: Evaluate the results using the same machine learning evaluation metrics as in Step 3 above.
[0086] This invention introduces physical modeling, making the prediction results more closely reflect the actual degradation process of batteries and improving interpretability. It also reduces computational complexity by optimizing feature inputs, enabling it to meet real-time or online prediction requirements. By combining the advantages of data-driven and physical modeling, this method achieves optimizations in reliability, accuracy, and applicability, providing an efficient and feasible solution for lithium-ion battery health management.
[0087] It should be noted and understood that various modifications and improvements can be made to the invention described in the detailed description above without departing from the spirit and scope of the invention as claimed in the appended claims. Therefore, the scope of the claimed technical solutions is not limited to any specific exemplary teachings given.
[0088] The applicant declares that the detailed method of the present invention is illustrated by the above embodiments, but the present invention is not limited to the above detailed method, that is, it does not mean that the present invention must rely on the above detailed method to be implemented. Those skilled in the art should understand that any improvements to the present invention, equivalent substitutions of the raw materials of the product of the present invention, addition of auxiliary components, selection of specific methods, etc., all fall within the protection scope and disclosure scope of the present invention.
Claims
1. A method for diagnosing the health status of lithium-ion batteries based on equivalent circuit feature screening, characterized in that, Includes the following steps: Step 1: Select graphite NCA 18650 battery Dataset_1 as the object. Its nominal capacity is 3.5 Ah, rated voltage is 3.6V, charging cut-off voltage is 4.2V, and discharging cut-off voltage is 2.65V. The battery dataset comes from the Zenodo database and contains charge and discharge data and frequency domain EIS electrochemical impedance spectroscopy data. The battery's charge and discharge protocol covers three rates: 0.5C, 1C, and 2C. The cycle temperature is set to a range of -20°C to 25°C. The specific test plan includes capacity calibration cycles, dynamic operating condition cycles, and EIS tests at different cycle intervals. Step 2: Preprocess the data and extract features; Step 3: Input the features into the machine learning model to perform the first assessment of the battery health status; Step 4: Establish an equivalent circuit model and extract relevant parameters, couple the parameters with ICA features, and further filter the features; Step 5: Input the filtered features to reassess the battery health status; Step 2 specifically includes: Step 2.1: The target value is the battery's health status data, represented as: SOH=Q n / Q0 Q n Q0 represents the capacity of the battery when it is currently fully charged, and Q0 represents the rated capacity of the battery. Step 2.2: Outlier handling: Use the Z-score method to filter out outliers. Data exceeding the threshold is replaced by the average of the previous and subsequent values. Step 2.3: Extract features from the charging data, including: constant current charging time, constant voltage charging time, initial voltage, average voltage, constant current charging time / constant voltage charging time, and constant current charging time / total charging time, thus obtaining 6 different features; Step 2.4: Process the charging data to obtain the incremental capacity curve; Step 2.5: If the peak disappears in the curve, find the point with the smallest absolute value of the derivative in the voltage window and take it as the value of the peak. Step 2.6: Select the peak value and peak voltage of each of the three peaks in the incremental capacity curve after processing the peak disappearance as the other 6 features to be input into the machine learning model. Step 3 specifically includes: Step 3.1: Input the 12 extracted features into the machine learning model; Step 3.2: Under different temperature conditions, set up different training and testing paths, and adopt a cross-validation strategy: at 25°C, use four batteries: 25C03, 25C06, 25C11, and 25C19, each as the test set, and the remaining batteries as the training set; and so on, at 35°C and 45°C, follow the same path, but with different batteries. At 35°C, use batteries 35C01 and 35C02; at 45°C, use batteries 45C13, 45C16, 45C21, and 45C27 respectively. This cross-validation framework effectively ensures the model's generalization ability to different individual batteries of the same model; the input is the 12 features obtained above, and the output is the battery health status; Step 3.3: Use commonly used machine learning evaluation metrics, including: mean squared error (MSE), root mean squared error (RMSE), and coefficient of determination (R²). Step 4 specifically includes: Step 4.1: Extract battery impedance data: including 100 data points in the frequency range of 50 mHz to 10 kHz, and extract once every 25 cycles; Step 4.2: Based on the obtained impedance data, use the impedance.py module of the Python library to establish and fit the equivalent circuit model, and construct a second-order RC equivalent circuit model. Step 4.3: Extract the parameters most relevant to battery degradation from the equivalent circuit model: R sei C sei C ct R ct Z w ; Step 4.4: Perform Pearson correlation analysis on the 12 features extracted in Step 2 and the 5 parameters in Step 4.
3. Sum the Pearson coefficients of each feature and the parameters, take the average, and sort them. Remove the features that are ranked lower or have an average value less than 0.6 to obtain the filtered features.
2. The method for diagnosing the health status of lithium-ion batteries based on equivalent circuit feature screening according to claim 1, characterized in that, Step 5 specifically includes: Step 5.1: Similar to the scheme in Step 3 above, a cross-validation strategy is adopted, with the input being the filtered features and the output being the battery health status; Step 5.2: Evaluate the results using the same machine learning evaluation metrics as in Step 3 above.