Electrochemical methods for battery quality assessment

Abstract

A method for identifying battery quality during battery formation includes: initiating life cycling after performing an initial battery formation charge on a plurality of batteries; collecting and preprocessing a discharge dataset generated by one of the plurality of batteries during the beginning of life cycling; calculating a statistical variance from the discharge dataset to determine an estimated probability of meeting a target battery life; and predicting the life of the plurality of batteries.

Description

Technical Field

[0001] The present disclosure relates to the manufacture of batteries. Background Art

[0002] The battery production methods and processes used to manufacture automotive battery cells typically include several testing and maintenance steps or practices. These include a first practice performed during cell formation, which defines a discharge capacity check to determine whether the cell delivers the predetermined ampere-hour (Ah) capacity. The measured Ah capacity must meet or exceed the specified value. Batteries that fail this first practice or step are typically set aside or discarded.

[0003] The second approach involves placing individual batteries in inventory for a minimum of 7-10 days and up to several months, during which time the open-circuit voltage of the batteries is monitored. The open-circuit voltage of the batteries is monitored during the inventory holding period to determine if the batteries have experienced "voltage sag," defined as a gradual decrease in open-circuit voltage over time. Batteries that exhibit a voltage sag exceeding a predetermined value or rate are identified as defective. The defective batteries are removed from the battery inventory and typically discarded. The battery storage capacity required to temporarily hold all batteries in production for 7-10 days or longer, as well as the cost of monitoring the battery voltage and storing the measurements, imposes undesirable costs and delays on the batteries and, therefore, on the battery pack manufacturing. Furthermore, the first and second approaches described above provide very limited diagnostic or pre-diagnostic capabilities in battery production, making it difficult to determine whether a battery has experienced a defect or requiring considerable time to determine whether the battery is of high, medium, or low quality.

[0004] The above process produces batteries with extremely high scrap rates and does not provide early manufacturing testing of battery quality. In addition, inventory retention time is too long and cannot be shortened by quality assurance earlier in the process. Battery accelerated life cycle testing is an aging and cycling test used to determine whether a specific batch of candidate batteries meets durability requirements based on 100 to 300 charge and discharge cycles. It is delayed during the time-consuming formation of the agreement due to the lack of quality control (QC) checks.

[0005] Therefore, while current battery manufacturing methods achieve their intended purposes, the production of automotive battery packs still requires a new and improved system and method for producing and testing batteries. Summary of the Invention

[0006] According to several aspects, a method for identifying battery quality during battery formation includes: initiating life cycling after performing an initial battery formation charge on a plurality of batteries; collecting and preprocessing a discharge dataset generated by one of the plurality of batteries during the beginning of life cycling; calculating a statistical variance from the charge and / or discharge dataset to determine an estimated probability of meeting a target battery life; and predicting the life of the plurality of batteries.

[0007] In another aspect of the present disclosure, the method further includes classifying the plurality of batteries into a short cycle-life group, a medium cycle-life group, and a long cycle-life group based on statistical variance.

[0008] In another aspect of the present disclosure, the method further includes applying peak detection to determine the peak location and magnitude of the voltage derivative of the resulting charge and beginning of life charge and discharge data curves, defined as dQ / dV curves.

[0009] In another aspect of the present disclosure, the method further includes determining whether humidity above a predetermined threshold value suppresses or shifts a peak on the dQ / dV curve due to reduced ethylene production.

[0010] In another aspect of the present disclosure, the method further includes correlating the various dQ / dV shape characteristics of the battery's charge / discharge curve with known long, medium, and short life groups of batteries to predict the expected life of the battery.

[0011] In another aspect of the present disclosure, the method further includes defining a cycle at the start of a life cycle operation, comprising: charging one of the plurality of batteries by increasing the battery voltage to approximately 4.2V; and discharging one of the plurality of batteries to reduce the battery voltage from approximately 4.2 to approximately 2.7V.

[0012] In another aspect of the present disclosure, the method further includes initiating life cycling operations for up to 10 cycles on each battery.

[0013] In another aspect of the present disclosure, the method further includes calculating a statistical variance of the voltage at a given charge level from a first cycle at the beginning of the life cycle to a maximum of 10 subsequent cycles.

[0014] In another aspect of the present disclosure, the method further includes converting the charge / discharge curve into a set of features including statistical variance, average battery charge and discharge values, and shape parameters, the shape parameters including deviations of the battery charge and discharge values, the battery charge and discharge values ​​including values ​​generated by left or right skewing from a suitable statistical distribution (including a Gaussian distribution), which are calculated based on the difference between the voltage or capacity of at least two cycles from the first cycle to at most the tenth cycle.

[0015] In another aspect of the present disclosure, the method further includes applying a predetermined threshold, wherein a single variance in the statistical variance or a variance above the predetermined threshold defines an out-of-specification battery in the plurality of batteries.

[0016] In another aspect of the present disclosure, the method further comprises constructing a battery cathode for a plurality of batteries having a cathode chemistry defined as LiNi x Mn y Co z O2(NMC622x≥0.6,y≤0.2,z≤0.2), LiMn a Fe (1-a) PO4(LMFP,a>0) and LiMn2O4(LMO) or a combination thereof.

[0017] According to several aspects, a method for identifying battery quality during battery formation includes: performing an initial battery formation charge on a plurality of batteries; collecting and preprocessing a formation charge data set generated by one of the plurality of batteries during the formation charge; smoothing the formation charge data set to remove noise; determining a derivative dQ / dV of a battery capacity (Q) with respect to a battery formation voltage (V) for the plurality of batteries; and performing peak fitting on a peak position of a data curve obtained by determining the derivative.

[0018] In another aspect of the present disclosure, the method further includes correlating factors including individual cell voltage, cell capacity, and conditions of individual additives among the plurality of additives added to the electrolyte in the cell.

[0019] In another aspect of the present disclosure, the method further includes applying different peak positions of the data curve during the battery formation charge to determine different ones of the plurality of conditions for individual batteries in the plurality of batteries.

[0020] In another aspect of the present disclosure, the method further includes determining an initial battery charge that occurs during formation of a single cell in the plurality of cells, and conducting the initial battery formation charge to a voltage of approximately 3.95V.

[0021] In another aspect of the present disclosure, the method further includes determining whether any of the plurality of cells was exposed to humidity above a predetermined threshold during formation.

[0022] In another aspect of the present disclosure, the method further includes constructing a plurality of containers, a single container receiving a condition of a battery of the plurality of batteries differentiated between a short cycle life group, a medium cycle life group, and a long cycle life group.

[0023] In another aspect of the present disclosure, the method further includes fitting each discharge voltage curve using a cubic spline or other suitable interpolation technique to obtain a set of voltage values ​​at a specific capacity or state of charge, and the voltage values ​​are generated using a set of capacity values ​​(Q) generated in increments of 4 mA-h between 0 and 1 A-hr. i ) (up to 250 steps); interpolate capacity and voltage data and calculate battery voltage at each Q increment; uniformly sample capacity and calculate battery voltage at a specific Q i Compare the voltages of adjacent cycles at the same value; for each capacity Q i , the difference between the second voltage curve and the first voltage curve is calculated to provide a set ξ, defined as and determine the statistical variance of each set ξ. Alternatively, the second set ξ' can be calculated in the same way, where Q is a function of voltage; ξ' = {Q2(V i )–Q1(V i ),1≤i≤250}.

[0024] According to several aspects, a method for identifying battery quality during battery formation includes: determining battery formation data for a single battery among a plurality of batteries during an initial formation charge event; determining battery discharge data for a single battery among a plurality of batteries during a start of life cycle events, for the first ten discharge events, preferably the first three discharge events, of the single battery among the plurality of batteries; combining the battery formation data and the battery start of life charge and / or discharge data with initial accelerated life cycle test data, and training pattern recognition on the battery formation data set; and predicting the battery life of the single battery among the plurality of batteries using the battery formation data set.

[0025] In another aspect of the present disclosure, the method further includes collecting battery discharge data using a battery voltage cycler with a voltage measurement accuracy of ≥ ±0.01% FSR, an accuracy of ±5 mV, and a range of w / 0-5V.

[0026] In another aspect of the present disclosure, the method further includes packaging the plurality of batteries into one of a first bin having high-quality batteries, a second bin having medium-quality batteries, and a third bin having low-quality batteries.

[0027] Further areas of applicability will become apparent from the description provided herein.It should be understood that the description and specific examples are intended for purposes of illustration only and are not intended to limit the scope of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings described herein are for illustration purposes only and are not intended to limit the scope of the present disclosure in any way.

[0029] Figure 1is a schematic diagram of a lithium-ion battery showing the cathode, anode, separator, positive current collector, and negative current collector;

[0030] Figure 2 Yes Figure 1 A chart of charging data is formed in;

[0031] Figure 3 yes Figure 2 The graph showing the derivative of capacity versus voltage dQ / dV in the charge curve is shown;

[0032] Figure 4 Yes Figure 1 The battery forms a graph of charging data versus the logarithm of capacity;

[0033] Figure 5 is a bar graph showing battery packs with different manufacturing defects according to an exemplary aspect

[0034] The remaining capacity of different batteries after 500 accelerated aging or life cycle tests;

[0035] Figure 6 is a graph showing the relationship between the remaining battery capacity, the logarithm of the change at the start of life discharge capacity and the voltage curve of the battery capacity of a battery produced according to one exemplary aspect;

[0036] Figure 7 is a dQ / dV graph showing the formation charge of batteries with different formation conditions;

[0037] Figure 8 is a flow chart of an automated process for forming a battery of the present disclosure;

[0038] Figure 9 is a flow chart of the entire process for preparing the battery of the present disclosure; and

[0039] Figure 10 are graphs detailing the voltage vs. time (top) and current vs. time for the initial formation charge and the beginning of the three lifetime charge / discharge cycles of the battery. DETAILED DESCRIPTION

[0040] The following description is merely exemplary in nature and is not intended to limit the present disclosure, application, or uses.

[0041] See Figure 1A system and electrochemical (EC) method for characterizing battery quality during the formation cycle 10 of an exemplary battery 12 is described. A cathode / separator / anode stack is constructed in a battery assembly. The initial battery assembly includes a separator 14, which can be a porous polymer material such as polypropylene or polyethylene, an anode 16, such as graphite, and a cathode 18, such as NMC662, positioned opposite the anode 16 through the separator 14. The stack of separator 14, anode 16, and cathode 18, along with the electrodes described below, is placed in a pouch 20. The battery 12 is initially a dry assembly of components and is not activated until the pouch 20 is filled with electrolyte 22. The three inner layers of the battery stack defining the anode 16, separator 14, and cathode 18 are porous. After the electrolyte is filled, the battery undergoes a "wetting process," which fills the pores with electrolyte 22. The battery 12 is considered functional when it reaches a battery activation charge with an open circuit voltage between approximately 0.1 VDC and approximately 2.0 VDC. Achieving the battery activation charge takes approximately two days, during which time data is collected and analyzed to obtain diagnostic data useful for battery qualification, as described below.

[0042] After confirming the battery activation, an initial battery formation charge is charged to approximately 3.95 V and applied to the battery. During the wetting period after the battery formation charge, the electrolyte 22 undergoes charge permeation, and the electrolyte solvent, additives, and salts are reduced on the outer surface of the anode 16 to form a first solid electrolyte film (SEI) 24 in situ on the surface of the anode 16. Simultaneously, during the wetting period, the electrolyte solvent, additives, and salts are oxidized on the surface of the cathode 18 to form a second solid electrolyte film (SEI) 26 in situ on the outer surface of the cathode 18.

[0043] An anode current collector 28, made of, for example, copper, is attached to the negative active material of the anode 16 and extends outward from the pouch 20. A cathode current collector 30, made of, for example, aluminum, is attached to the positive active material of the cathode 18 and extends outward from the pouch 20. The reduction of the electrolyte 22 and any contaminants produces different electrochemical reactions. The formation of the first SEI 24 and the second SEI 26 is accomplished by reducing the electrolyte 22, which defines a variety of electrolyte solvents, additives, and salts, all at a specific voltage. The reduction of the electrolyte 22 also emits a variety of formed gases 32, which can be collected in separate areas of the pouch 20 and discharged from the pouch 20.

[0044] Electrolyte penetration of the separator 14, the active material of the anode 16, and the active material of the cathode 18 is defined as "wetting."

[0045] Due to the voltage difference, or if current is applied from cathode 18 to anode 16 and vice versa, the ions migrate spontaneously, which takes about two days. It has been determined that after the electrolyte 22 is introduced and the charge is initiated, the changes in each electrochemical response curve are proportional to the amount of decomposition within a specific voltage range.

[0046] The active materials commonly used in lithium-ion batteries are as follows:

[0047] cathode: LiNi x Mn y Co z O2 (NMC622x≥0.6, y≤0.2, z≤0.2),

[0048] LiMn a Fe (1-a) PO4(LMFP, a>0), LiMn2O4(LMO), or mixture

[0049] Anode: Lithium ion: Anode is graphite: SiOx, Si or mixture

[0050] Lithium metal: the anode is lithium metal

[0051] Gases produced when SEI forms include:

[0052] C2H4, CO, H2, CH4, C2H6, butane, etc.

[0053] See Figure 2 And refer again Figure 1 , the formed charge graph 38 of the battery 12 provides an exemplary charge curve 40 of measured voltage 42 versus time compared to measured battery capacity (Ah) 44. The charge curve 40 is substantially vertical before reaching an inflection region 46. The data collected from the charge curve 40, particularly the inflection region 46, is analyzed to identify the quality level of the individual cells of the battery 12, which will be referenced. Figure 3-5 Detailed description.

[0054] See Figure 3 And refer again Figure 2 , using an exemplary cell chemistry of NCM622 / graphite cell 12,

[0055] Graph 48 illustrates a method for analyzing the charge formation by comparing the derivative dQ / dV of the battery capacity (Q) with respect to the battery formation voltage (V) indicated on a first axis 50 and the formation voltage (V) indicated on a second axis 52. Figure 1The cell formation is described when there is no additive in the electrolyte 22. The cell has almost no response until a voltage spike occurs at approximately 2.9 volts. A cell with this response is expected to have the shortest life and may not be acceptable. The second curve 56 represents the cell formation when there is a good additive in the electrolyte 22. Cells with this response are expected to be of the best or very high quality and have the longest life. The third curve 58 represents the cell formation when there is an aged additive in the electrolyte 22. Cells with this response are expected to be of medium quality and have a medium life. The fourth curve 60 represents the cell formation when the cell is exposed to high ambient humidity above a predetermined humidity. Cells with this response are expected to be of poor quality and have a medium to shortest life. As can be seen from the above, the shape of the charge curve may be related to a variety of factors, including the condition of the additive and the presence or absence of high humidity, and to the expected life of the battery.

[0056] Figure 3 The data presented in the present invention allows the identification of whether the cell quality is high, medium or low quality early in the formation of the cell and before accelerated life cycle testing. This cell quality pattern recognition allows the cells to be sorted into high quality groups, medium quality groups and low quality groups after initial formation. The sorting can be done by assigning different cell masses to the individual cells. Since the use of all cells with similar cell masses is beneficial for the subsequent battery pack assembly, Figure 3 The data in allows for a battery pack with the highest overall quality and longest potential life.

[0057] See Figure 4 And refer again Figure 2 and 3 , a graph 62 presents the battery formation data in different formats, wherein a first axis 64 defines voltage and a second axis 66 defines the logarithm of battery capacity (Ah). Different regions 68 determine when the battery formation data curves differ. A first curve 70 defines battery formation without additives in the electrolyte. A second curve 72 defines battery formation under dry (low humidity) conditions with appropriate additives. A third curve 74 defines battery formation that occurs mostly under humidity conditions above a predetermined threshold.

[0058] See Figure 5 And refer again Figures 1 to 4 , histogram 76 shows Figure 4 The remaining capacity (%) of the three batteries after 500 cycles of the accelerated life cycle test is used to distinguish acceptable batteries from defective batteries. The first bar 80 represents Figure 4 The first curve 70 in FIG. 1 identifies a battery having a remaining capacity of approximately 70% as a defective battery. The second bar 82 represents Figure 4 The second curve 72 in FIG. 1 identifies a battery having a remaining capacity of approximately 92% as a high-quality battery. The third bar 84 represents Figure 4 The battery defined by the third curve 74 in FIG. 1 is identified as an acceptable medium battery with a remaining capacity of approximately 87%. As can be seen from the above, the variation in the discharge curve may be related to various factors including the battery voltage and the humidity during battery formation, and the battery capacity is related to the expected life of the battery.

[0059] Pattern recognition in cell formation cycling data is combined with limited accelerated life cycle testing to create learning feedback, which can narrow or completely eliminate the time window for conducting accelerated life cycle testing. Feedback identified in cell formation cycling is identified to provide more timely corrective actions in battery production. Explicit quality checks early in the production process reduce the need for voltage drop testing of cells and battery packs during storage. Data-rich process monitoring improves battery quality and is more effective in assembly rate-limiting steps. Data processing using advanced analytics is used to generate and monitor key features of the electrochemical signature.

[0060] Equation 1:

[0061]

[0062] Where: s 2 = Sample variance

[0063] x=x i = the value of the i-th element, i = 1...n

[0064] xbar = sample mean

[0065] n = sample size

[0066] It has been found that by calculating the statistical variance S from the first 3 (three) cycles using the above equation 1, it is possible to provide an estimated probability of meeting the target usage time, allowing the batteries to be classified into short expected cycle life groups, medium expected cycle life groups, and long expected cycle life groups, thereby reducing the reliance on accelerated life cycle testing. For example, in a first step, cubic spline interpolation is used to fit each discharge voltage curve to obtain a set of voltage values ​​at a specific capacity or state of charge. To this end, a set of capacity values ​​(Q) with increments of 4 mAh are generated between 0 and 1 Ah (250 steps). i ), and then use a cubic spline fit to the experimental capacity vs. voltage data to calculate the corresponding voltage at each increment Q. Uniform capacity sampling allows i For each capacity Q i , the difference between the second voltage curve and the first voltage curve is calculated to provide a set ξ, defined as It can be abbreviated as ΔV 2-1For each set ξ, take the variance as shown in Equation 1. Alternatively, the second set ξ' can be calculated in the same way, where Q is a function of voltage;

[0067] See Figure 6 And refer again Figures 1 to 5 By calculating the statistical variance from the first three cycles of battery operation using Equation 1 above, the probability of meeting the target usage time can be estimated. Using this variance, the batteries can be grouped into a short cycle life group or low quality group, a medium cycle life group or medium quality group, and a long cycle life group or high quality group. Figure 6 The graph 86 shown has the remaining battery capacity (%) at 500 cycles as the first axis 88. The second axis 90 defines the logarithm of the calculated variance. The battery shown in the first region 92 can be defined as a battery with a long cycle life. The battery shown in the second region 94 can be defined as a battery with a medium cycle life. The battery shown in the third region 96 can be defined as a battery with a short cycle life. As can be seen from the above, the change in the discharge curve may be related to various factors including the battery capacity, and the change in battery capacity is related to the expected life of the battery.

[0068] The data collected during the cell formation phase are automatically preprocessed using noise filtering to smooth the data. Then, for example, first-order or higher-order derivatives such as Figure 3 The data is converted to dQ / dV as described in . Peak detection is then used to identify features to determine peak positions and intensities.

[0069] See Figure 7 And refer again Figures 2 to 6 , the data shown in graph 98 corresponds to the derivative (dQ / dV) of the cell capacity (Q) with respect to the cell formation voltage (V) identified on a first axis 100 and the formation voltage (V) on a second axis 102. Peaks in the data correspond to reduction of solvents, salts, or additives in the electrolyte 22. For example, a large peak 104 in dQ / dV corresponds to a lack of additives in the electrolyte 22. The presence of moisture suppresses the EC peak 106 due to, for example, reduced ethylene production from exposure of the cell to moisture during formation. Increasing moisture causes a proportional increase in the intensity of the reduction feature and suppresses EC reduction. These deviations can be correlated to formed gases and combined to determine the root cause of the defect.

[0070] See Figure 8 And refer again Figures 2 to 7 , an exemplary automatic or automated process flow diagram 108 includes an initial or first step 110 of determining an initial battery charge during battery formation. Figures 2 to 4 In a second step 112, during the battery formation, reference is made to Figure 6 The life cycle is started. The data of the charging process is collected and the data is pre-processed using noise filtering to smooth the data. Then refer to Figure 3 and 7 The data is converted using derivatives such as dQ / dV. For example, features are identified by peak detection to determine peak position and intensity, such as Figure 6 In the third step 114, if necessary, inventory preservation and initial accelerated life cycle testing can be performed based on the results of the first step 110 and the second step 112.

[0071] See Figure 9 And refer again Figures 2 to 8 , an exemplary system flow chart 116 identifies the characteristics of the system and the electrochemical methods used to identify battery quality during battery formation 10. The following processes or steps are all automated. The server computer 118 collects all system data from the repository 120 and makes the final decision on the battery quality, including determining whether the battery is of high quality 122, medium quality 124 or low quality 126, and further identifying unacceptable or poor quality batteries. The repository 120 receives all sensor data and decisions from the edge computer. The server computer 188 also provides communication of new system "rules" and communicates these rules to the edge computer 128. The edge computer 128, which can be a single computer or multiple computers, provides sensing and single-station monitoring. The edge computer 128 is connected to at least one sensor for each battery being prepared. The server computer 118, the repository 120, and the edge computer 128 belong to the same network. Each computer includes one or more processors, memory, and instructions stored in the memory. Memory is a non-transitory computer-readable medium.

[0072] A plurality of materials, including a solvent 130, a binder 132, an active material 134, and carbon black 136, are combined to produce a first slurry 138. First slurry 138 can be combined with aluminum 140 to produce a cathode 142, similar to cathode 18 described above. Edge computer 128 monitors the components forming cathode 142 and the ambient temperature and humidity conditions during the formation of cathode 142. Cathode 142, separator 144, and anode 148 are combined to partially form battery assembly 146, with anode 148 being similar to anode 16 described above, and the entire assembly being further defined below. A plurality of materials, including a solvent 150, a binder 152, and an active material 154, are combined to produce a second slurry 156. Copper 158 can be combined with second slurry 156 to produce anode 148, similar to anode 16 described above. Edge computer 128 monitors the components forming anode 148 and the ambient temperature and humidity conditions during the formation of anode 148.

[0073] After the battery assembly 146 is assembled, the electrolyte 22 is added and a wetting process 160 is performed, which is monitored by the edge computer 128. The wetting process 160 is followed by a cell formation process 162, which is independently monitored by the edge computer 128. The cell formation process 162 is followed by a degassing process 164, which is independently monitored by the edge computer 128. Finally, after the degassing process 164 is completed, a life cycle process 166 begins, which is independently monitored by the edge computer 128.

[0074] The time of each step in the above process is recorded by the edge computer 128 and transmitted to the server 118.

[0075] See Figure 10 Graph 168 shows the battery discharge data of the battery voltage 170 and battery current 172 (in milliamperes) to time 174 (in hours) for a single battery in the present application at the beginning of a life cycle test. Discharge data is initially collected at the beginning of three life charge / discharge cycles 176, 178, and 180. A cubic spline or similar fit is used to fit the data for each cycle, and the capacities are aligned by interpolation. The voltage change between each two cycles is then calculated, for example, given as ΔV 2-1 . ΔV is then calculated using Equation 1 above 2-1 The calculated statistical variance is related to the previously performed accelerated batch acceptance test.

[0076] The subtle electrochemical reactions that occur during a battery's charging cycle reveal quality issues. Pattern recognition, using statistical data and machine learning, is used to identify quality issues, allowing defective batteries to be identified early in production, before accelerated life cycle testing. Good batteries can also be identified and categorized as low-quality, medium-quality, or high-quality.

[0077] To collect battery voltage discharge data, a circulator with a voltage measurement accuracy of ≥ ±0.01% full-scale range (FSR) (e.g., ±5 mA in the 0-5 V range) can be used. These lower-accuracy circulators are less expensive than known higher-accuracy circulators, which have a current measurement accuracy of ≥ ±0.02% FSR (e.g., ±10 mA in the 0-5 V range) and a current control resolution of 0.0003% FSR.

[0078] In another aspect of the present disclosure, the method further includes converting the charge / discharge curve into a set of features including statistical variance, average battery charge and discharge values, and shape parameters (deviations) of the battery charge and discharge values, wherein the battery charge and discharge values ​​include values ​​generated by left or right skewing from a suitable statistical distribution (e.g., a Gaussian distribution), which are calculated based on the difference between the voltage or capacity of at least two cycles from the first cycle to a maximum of the tenth cycle.

[0079] The disclosed system and electrochemical method for identifying battery quality during battery formation 10 provides several advantages. These advantages include a method that combines electrochemical characterization of batteries with statistical and machine learning data to identify quality issues that may arise during production. Utilizing data from formation and the beginning of life cycling, combined with initial accelerated cycling tests, and using feedback from these tests to train a pattern recognition algorithm for formation responses, can lead to the gradual elimination of accelerated cycling tests and provide production feedback earlier in battery production.

[0080] The description of the present disclosure is merely exemplary in nature, and variations that do not depart from the gist of the present disclosure are intended to fall within the scope of the present disclosure. Such variations should not be regarded as departing from the spirit and scope of the present disclosure.

Claims

1. A method for identifying battery quality during battery formation, comprising: Initiating life cycling after performing an initial battery formation charge on the plurality of cells; collecting and preprocessing a discharge dataset generated by one of the plurality of batteries during a beginning of life cycle; calculating a statistical variance from the discharge data set to determine an estimated probability of meeting a target battery life; predicting lifespans of the plurality of batteries; Defines the cycle at which the life cycle operation begins, including: charging one of the plurality of batteries by increasing the battery voltage to 4.2V; and discharging one of the batteries to reduce the battery voltage from 4.2V to 2.7V; The life cycle operation is started only for the first, second and third cycles; calculating the statistical variance from the first cycle, the second cycle, and the third cycle at the start of life cycles; and The discharge dataset is converted into a set of discharge feature sets including the statistical variance, the average battery charge and discharge values, and shape parameters, wherein the shape parameters include deviations of battery charge / discharge values, the battery charge / discharge values ​​including values ​​resulting from a left or right skew from a statistical distribution including a Gaussian distribution, and the values ​​are calculated based on differences between voltages or capacities of at least two cycles from a first cycle to a tenth cycle. 2 . The method of claim 1 , further comprising classifying the plurality of batteries into a short cycle life group, a medium cycle life group, and a long cycle life group based on the statistical variance.

3. The method of claim 2 , further comprising determining whether humidity above a predetermined threshold suppresses peak formation in a discharge data curve due to reduced ethylene production in the plurality of cells, and correlating the statistical variance with a shortened predicted life of a plurality of cells between one of the medium cycle life group and the short cycle life group. 4 . The method of claim 1 , further comprising applying a predetermined threshold, wherein a single one of the statistical variances that is above the predetermined threshold defines an out-of-specification battery in the plurality of batteries.

5. The method of claim 1 further comprising constructing battery cathodes for said plurality of batteries having a cathode chemistry defined as LiNi x Mn y Co z O2 (NMC622 x≥0.6, y≤0.2, z≤0.2), LiMn a Fe (1-a) One of PO4 (LMFP, a>0) and LiMn2O4 (LMO) or a combination thereof.

6. The method of claim 5, further comprising constructing battery anodes of graphite material for the plurality of batteries having an anode chemistry defined as one of SiOx or Si.

Images