Method for controlling early performance of battery cell
By analyzing the QV and dQ/dV curves of the cells during the lithium-ion battery formation stage, characteristic peaks were identified and characteristic vectors were constructed, solving the cell inconsistency problem, realizing early fine sorting, improving the stability and safety of the battery pack, and reducing costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGMEN ZETA POWER SUPPLY TECH CO LTD
- Filing Date
- 2026-03-24
- Publication Date
- 2026-07-28
AI Technical Summary
In existing lithium-ion battery packs, performance differences between cells lead to voltage imbalances and inconsistencies, affecting the stability and safety of the battery pack. Furthermore, traditional sorting methods are outdated and cannot identify microscopic differences, resulting in economic losses and safety risks.
By acquiring voltage and current data at high frequency during the cell formation stage, QV charging curves and differential capacity dQ/dV curves are generated, main characteristic peaks are identified, feature vectors are constructed, and Gaussian mixture model clustering and support vector regression models are used for sorting to screen out inconsistent cells, and accelerated life testing and safety warnings are conducted.
It enables early and refined control over the initial state of the battery cell, identifies inconsistent cells in advance, improves the stability and safety of the battery pack, reduces costs, and delays performance degradation.
Smart Images
Figure CN122474739A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of lithium-ion battery management, specifically relating to a method for controlling the early performance of a battery cell. Background Technology
[0002] In lithium-ion battery packs, significant performance differences between cells can lead to voltage imbalances and inconsistencies during discharge. Voltage imbalances reduce the effective capacity of the battery pack, impacting its range. Furthermore, cell inconsistencies can cause heat buildup within the battery pack, severely affecting its stability and safety. Therefore, cell consistency is crucial to the overall performance of lithium-ion battery packs, and ensuring this consistency as early as possible has become a significant technical challenge in the industry. Summary of the Invention
[0003] The purpose of this invention is to provide a method for controlling the early performance of battery cells, which addresses the shortcomings of existing technologies and can significantly improve the control accuracy of battery cells and battery packs, and prevent inconsistencies in battery cells in the early stages.
[0004] To achieve the above objectives, the present invention adopts the following technical solution: A method for controlling the early performance of a battery cell includes the following steps: Step 1: During the formation stage of at least 10 cells from the same batch, the voltage and current of each cell are synchronously collected and recorded at a sampling frequency of at least 10Hz. The real-time cumulative capacity is calculated by integration to generate the QV charging curve for each cell. The actual real-time measured data is smoothed and denoised. Then, the differential capacity dQ / dV of each cell is calculated to generate the QV charging curve for each cell. -V curve; Step 2: For each battery cell -V curve, automatically recognized The main characteristic peaks are identified, and the peak voltage Vp, peak intensity Ip, peak area Ap, and full width at half maximum (Wp) are extracted. Based on the corresponding characteristic parameters, a feature vector representing the initial state of each cell is constructed. Step 3: Perform sorting and performance evaluation management to screen out cells that do not meet the requirements.
[0005] As an improvement to the method for controlling the early performance of the battery cells described in this invention, step one further includes: employing [a specific method] during the formation and charging process of each battery cell. Constant current charging to , transformed into It is carried out in an environment.
[0006] As an improvement to the method for controlling the early performance of the battery cells of the present invention, step one further includes: each of the selected battery cells is a lithium iron phosphate-graphite system battery cell.
[0007] As an improvement to the method for controlling the early performance of the battery cell described in this invention, step two further includes: identifying the main characteristic peak through a peak detection algorithm containing continuous wavelet transform, and further extracting the characteristic parameter of peak skewness S to characterize the initial state of the battery cell.
[0008] As an improvement to the method for controlling the early performance of the battery cell described in this invention, step three further includes: using a Gaussian mixture model clustering algorithm to iteratively calculate the feature vectors from step two to obtain grouping results.
[0009] As an improvement to the method for controlling the early performance of the battery cell described in this invention, step three further includes: after standard capacity testing and aging of the battery cell, conducting accelerated cycle life testing, with test conditions of 1C charging, 1C discharging, and 100% depth of discharge, recording the capacity retention rate after several cycles, and using the capacity retention rate as the output label to train a support vector regression model.
[0010] As an improvement to the method for controlling the early performance of the battery cells described in this invention, step three further includes: recording the total formation and charging capacity Q_formation of each battery cell, plotting a scatter plot with Q_formation as the abscissa and peak intensity Ip as the ordinate, and screening out abnormal battery cells based on the internal quality differences of the battery cells and issuing a safety warning.
[0011] As an improvement to the method for controlling the early performance of the battery cells described in this invention, it further includes step four: constructing a battery pack whose cell consistency meets the required conditions, performing dQ / dV analysis on all cells in the battery pack, extracting feature vectors F=[Vp,Ip,Ap,Wp], calculating the Euclidean distance D_ij between the feature vectors of every two cells to quantify the initial state difference, forming a difference matrix, and then selecting the cells with the smallest maximum difference D_max in the battery pack for combination as needed.
[0012] The beneficial effects of this invention are as follows: 1) By analyzing the differential capacity curves in a specific low-voltage range, the differential capacity analysis dQ / dV can sensitively resolve the electrochemical phase transition and reaction kinetics processes inside the battery, realizing early, non-destructive, and refined performance sorting control and prediction of the initial state inside the cell, and can more accurately predict the long-term performance of the cell; 2) This invention can identify and ensure the consistency of the cell one manufacturing stage earlier than traditional capacity sorting, advance the sorting node, realize earlier quality control and cost savings, and identify potentially inconsistent cells before investing in subsequent packaging and capacity sorting costs. Attached Figure Description
[0013] Figure 1 This is a schematic diagram of the process of the present invention.
[0014] Figure 2 This is a statistical chart showing the differential situation of different cells from the same batch of the present invention due to their own differences. Detailed Implementation
[0015] Specifically, cell consistency refers to the degree of similarity in performance parameters among individual cells within the same battery pack. These parameters can include voltage, capacity, internal resistance, and self-discharge rate. For battery pack performance, the quality of cell consistency directly affects the overall performance, safety, and lifespan of the battery pack. The inventors discovered that existing cell consistency sorting operations are mainly performed after capacity grading or capacity testing, based on the macroscopic performance of the cells (such as total capacity and internal resistance). The inventors found that this "post-sorting" has significant drawbacks: 1) It has a large lag, as sorting occurs after significant manufacturing costs (such as electrolyte, labor, and time) have already been invested, and economic losses due to defective products have already occurred; 2) It only reflects superficial characteristics and cannot distinguish cells with the same macroscopic capacity but different initial microscopic states (such as SEI film quality and initial lithium intercalation uniformity of active materials). These microscopic differences are the root cause of long-term cycle performance differentiation. Therefore, traditional formation process monitoring typically only focuses on total current, voltage, and final capacity, lacking a detailed analysis of the complex electrochemical reactions during the process, and thus cannot analyze the cell consistency situation. Therefore, a new technological approach is urgently needed to solve the above problems.
[0016] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following description is provided in conjunction with the accompanying drawings. The following is a detailed description of the specific implementation methods, structures, features, and effects of the present invention, as well as preferred embodiments.
[0017] A method for controlling the early performance of a battery cell, as shown in the figure. As shown, it includes the following steps: Step 1: During the formation stage of at least 10 cells in the same batch, synchronously collect and record the voltage and current of each cell at a sampling frequency of at least 10Hz. Calculate the real-time cumulative capacity through integration to generate the QV charging curve for each cell. Filter and smooth the actual real-time measured data to remove noise. Then calculate the differential capacity dQ / dV for each cell to generate the QV charging curve for each cell. -V curve; Step 2: For each battery cell -V curve, automatically recognized The main characteristic peaks are identified, and the peak voltage Vp, peak intensity Ip, peak area Ap, and full width at half maximum (Wp) are extracted. Based on the corresponding characteristic parameters, a feature vector representing the initial state of each cell is constructed. Step 3: Perform sorting and performance evaluation management to screen out cells that do not meet the requirements.
[0018] in, The reactions within the voltage range are typically closely related to the film formation kinetics of the negative electrode SEI film, the initial reduction of the electrolyte, and the early lithium intercalation behavior of the active material, containing crucial information about the cell's inherent properties. Simultaneously, the extracted characteristic parameters—peak voltage Vp, peak intensity Ip, peak area Ap, and full width at half maximum (WHM)—directly reflect the quality of initial reactions such as SEI film formation, closely related to the cell's inherent properties, and their predictive ability for long-term cycle life and storage performance far surpasses that of traditional macroscopic parameters. Furthermore, batch-wise shifts in characteristic parameters (such as a systematic leftward shift of peak voltage Vp and a general decrease in peak area Ap) can directly and quickly pinpoint upstream process issues (such as negative electrode slurry dispersion, electrolyte formulation, or fluctuations in injection volume), providing high-value real-time data for process closed-loop optimization and effectively guiding process optimization.
[0019] Preferably, step one further includes: each selected battery cell is a lithium iron phosphate-graphite system battery cell, and each battery cell adopts the following during the formation and charging process: Constant current charging to .
[0020] Preferably, step three further includes: using a Gaussian mixture model clustering algorithm to iteratively calculate the feature vectors from step two to obtain the grouping results.
[0021] Preferably, step three also includes: after standard capacity testing and aging of the battery cells, conducting accelerated cycle life testing, with test conditions of 1C charging, 1C discharging, and 100% depth of discharge, recording the capacity retention rate after several cycles, using the capacity retention rate as the output label, and training a support vector regression model.
[0022] Preferably, step three also includes: recording the total formation and charging capacity Q_formation of each cell, plotting a scatter plot with Q_formation as the abscissa and peak intensity Ip as the ordinate, and screening out abnormal cells based on the internal quality differences of the cells and issuing safety warnings.
[0023] Preferably, step four further includes: constructing a battery pack whose cell consistency meets the required conditions, performing dQ / dV analysis on all cells in the battery pack, extracting feature vectors F=[Vp,Ip,Ap,Wp], calculating the Euclidean distance D_ij between the feature vectors of every two cells to quantify the initial state difference, forming a difference matrix, and then selecting the cells with the smallest maximum difference D_max in the battery pack for combination as needed.
[0024] Example 1: Early clustering based on dQ / dV characteristics during formation. 3000 lithium iron phosphate cells from the same batch were selected and subjected to formation charging at 25℃ (using a 0.02C constant current to 3.0V). A data acquisition system synchronously recorded voltage and current at a frequency of 50Hz. For each cell's formation charging data, dQ / dV curves were calculated and generated. A peak detection algorithm with continuous wavelet transform was used to automatically identify the main characteristic peak in the 2.5V-2.7V range, extracting peak voltage Vp, peak intensity Ip, and peak area Ap. Using a Gaussian mixture model (GMM) clustering algorithm, with the above three features as input, the cells were finally divided into three groups: A (high-quality, 75%), B (good, 20%), and C (concerned, 5%). Results: Group A showed sharp peaks and concentrated Ap; Group B had slightly smaller Ap or slightly shifted Vp; Group C had significantly smaller Ap and a broadened peak, corresponding to batches where the humidity of the electrolyte injection process was out of control. Through appropriate operations, this sorting process can be completed at least 48 hours earlier than traditional capacity sorting.
[0025] Example 2: Early grading of battery cell cycle life prediction. 500 cells are randomly selected from the production line, and the above steps are performed to obtain dQ / dV characteristic parameters. Then, after standard capacity testing and aging, accelerated cycle life testing (1C / 1C, 100% DOD, 45℃) is conducted, and the capacity retention rate after 500 cycles is recorded. Using peak area Ap and full width at half maximum (Wp) as input features and capacity retention rate as the output label, a support vector regression model is trained. Application: The lifespan grade of newly produced cells can be predicted using the dQ / dV characteristics from the formation stage—retention rate >92% is classified as "long-life group" (automotive), 88%-92% as "standard group" (energy storage), and <88% as "degraded group."
[0026] Example 3: Composite Sorting Combining Formation Capacity and Differential Characteristics. After formation and charging, record the total formation and charging capacity Q_formation for each cell, and simultaneously obtain the dQ / dV characteristic peak intensity Ip in the 2.5V-2.7V range. Plot a scatter plot with Q_formation as the x-axis and Ip as the y-axis, sorting by quadrant: ① High Q, High Ip (optimal, high activity and good reaction kinetics); ② High Q, Low Ip (needs attention, uneven reaction); ③ Low Q, High Ip (abnormal, micro-short circuit or abnormal lithium consumption); ④ Low Q, Low Ip (worst). Effect: This method can precisely distinguish the internal quality differences of traditional "qualified capacity" cells, quickly detect abnormal cells, and trigger alarms for necessary safety warnings.
[0027] Example 4: Optimal Consistency Allocation of High-End Battery Packs. From 2000 qualified cells, 240 cells (20 series, 12 parallel) are selected to form a high-end energy storage battery pack. All cells undergo dQ / dV analysis to extract the feature vector F=[Vp,Ip,Ap,Wp]. The Euclidean distance D_ij between the feature vectors of every two cells is calculated (quantifying the initial state difference), forming a difference matrix. An optimization algorithm is used to select the 240 cell combinations with the smallest maximum difference D_max within the pack. Comparative verification: After 300 cycles, the voltage range increase of the battery pack optimized in this invention is reduced by approximately 60% compared to traditional capacity-graded allocation, significantly delaying performance discrepancies.
[0028] Where, when the feature vector of cell i is The eigenvector of cell j is At that time, the Euclidean distance between the eigenvectors of the two battery cells .
[0029] This solution sets up a sorting node after cell formation. The principle of this invention is also applicable to the characteristic analysis of other lithium-ion battery systems during the formation stage, and can significantly improve the control accuracy of cells and battery packs. This solution overcomes the shortcomings of existing technologies and provides a predictive early-stage cell sorting and control technology based on low-voltage range differential capacity analysis that can be implemented in the formation process.
[0030] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.
Claims
1. A method for controlling the early performance of a battery cell, characterized in that, Includes the following steps: Step 1: During the formation stage of at least 10 cells in the same batch, synchronously collect and record the voltage and current of each cell at a sampling frequency of at least 10Hz. Calculate the real-time cumulative capacity through integration to generate the QV charging curve for each cell. Smooth and denoise the actual real-time measured data, and then calculate the differential capacity dQ / dV for each cell to generate the QV charging curve for each cell. -V curve; Step 2: For each battery cell -V curve, automatically recognized The main characteristic peaks are identified, and the peak voltage Vp, peak intensity Ip, peak area Ap, and full width at half maximum (Wp) are extracted. Based on the corresponding characteristic parameters, a feature vector representing the initial state of each cell is constructed. Step 3: Perform sorting and performance evaluation management to screen out cells that do not meet the requirements.
2. The method for controlling the early performance of a battery cell as described in claim 1, characterized in that, Step one also includes: using the formation and charging process of each battery cell. Constant current charging to .
3. The method for controlling the early performance of a battery cell as described in claim 1, characterized in that, Step one also includes: each of the selected cells is a lithium iron phosphate-graphite system cell.
4. As claimed The method for controlling the early performance of a battery cell according to any one of the claims is characterized in that, Step two also includes: identifying the main feature peak using a peak detection algorithm containing continuous wavelet transform.
5. As claimed in the following claims The method for controlling the early performance of a battery cell according to any one of the claims is characterized in that, Step three also includes: using a Gaussian mixture model clustering algorithm to iteratively calculate the feature vectors from step two to obtain the grouping results.
6. As claimed The method for controlling the early performance of a battery cell according to any one of the claims is characterized in that, Step three also includes: after standard capacity testing and aging of the battery cells, accelerated cycle life testing is carried out. The test conditions are 1C charging, 1C discharging, and 100% depth of discharge. The capacity retention rate is recorded after several cycles, and the capacity retention rate is used as the output label to train the support vector regression model.
7. As claimed The method for controlling the early performance of a battery cell according to any one of the claims is characterized in that, Step three also includes: recording the total formation and charging capacity Q_formation of each cell, plotting a scatter plot with Q_formation as the abscissa and peak intensity Ip as the ordinate, and screening out abnormal cells based on the internal quality differences of the cells and issuing safety warnings.
8. As claimed in the following claims The method for controlling the early performance of a battery cell according to any one of the claims is characterized in that, The process also includes step four: constructing a battery pack with cell consistency that meets the required conditions, performing dQ / dV analysis on all cells in the battery pack, extracting feature vectors F=[Vp,Ip,Ap,Wp], calculating the Euclidean distance D_ij between the feature vectors of every two cells to quantify the initial state difference, forming a difference matrix, and then selecting the cells with the smallest maximum difference D_max in the battery pack for combination as needed.