Method for Rapid Screening of Rechargeable Batteries

Abstract

Old batteries are screened based on the measured Specific Charge Current Ratio (SCCIR). The old battery is partially charged at a constant current (CC) within a small voltage range until a voltage target is reached, and the current is integrated to obtain the applied CC charge Qcc. Then, the battery is continued to be charged at a constant voltage (CV) with the voltage target until the charging current drops to a mid-range current target without reaching full charge. The current during the CV period is integrated to obtain the applied CV charge Qcv. The measured SCCIR is the ratio of Qcc to (Qcc + Qcv). The measured SCCIR is input into a calibration curve function to obtain a model State of Health (SOH) value for classification. By aging new batteries, SCCIR and SOH data are obtained and modeled using a neural network to obtain the calibration curve function.

Description

[0001] Related Applications

[0002] This application is a Continuation-In-Part (CIP) of the patent application "Method for Rapid Screening of Old Batteries Calibrated with Constant-Current Impulse Ratio (CCIR)" with U.S. Patent Application No. 17 / 169,675, filed on February 8, 2021. Technical Field

[0003] The present invention relates to a battery screening method, and particularly to a method for screening used or retired batteries for reuse. Background Art

[0004] Batteries are widely used to power various systems, including electric vehicles (EVs). Each EV requires a large battery pack to provide the substantial power needed to propel the EV.

[0005] EVs typically use relatively expensive lithium-ion batteries. The chemicals used in such advanced batteries pose disposal problems. Toxic chemicals can leak from discarded batteries and contaminate water sources. With the popularization of EVs, as EV batteries are retired, it will impose an additional burden on landfills.

[0006] Recycling lithium-ion batteries and other batteries may require the use of acids or blast furnaces, which can cause other environmental problems. Low profit margins make battery recycling unattractive.

[0007] In particular, EV battery packs may be replaced prematurely. Recommendations from EV manufacturers may require repair shops to replace battery packs with a relatively high discharge capacity below that required to ensure the performance of the EV. EV battery packs may be removed before all battery cells are depleted. Especially for large battery packs, there may be many battery cells or battery packs that still have a considerable remaining service life. These battery cells can be used to power other systems with less stringent power requirements, such as communication and computer backup systems. Instead of discarding the replaced EV batteries in landfills or melting them, reusing EV batteries and extending their service life by 5 to 7 years provides a more sustainable and environmentally friendly approach.

[0008] The availability of used batteries can be defined by their State-of-Health (SOH) ratio. SOH refers to the ratio of the current storage capacity of a battery to its initial or nominal storage capacity. The storage capacity is approximately the discharge capacity of the battery.

[0009] Figure 1 Shows a prior art battery capacity test. There can be many variations, Figure 1For illustration only and does not necessarily represent any specific battery test.

[0010] Accurately measuring the full storage capacity of a battery can take a long time. Fast charging or discharging can heat the battery and affect the measurement. The battery may initially store some remaining charge that needs to be discharged before capacity measurement.

[0011] In step 202, first charge the battery under test to 3.8 volts by applying a constant current (CC) of 1C ampere value, and then once the target voltage of 3.8 volts is reached, reduce the current to maintain a constant voltage (CV) or 3.8 volts. The current will decrease during the CV phase until a low current value, such as 0.01C, is reached or until a certain period of time has elapsed.

[0012] Let the battery cool for one hour before proceeding to the next step. Similarly, the battery can be allowed to cool for 10 minutes before the initial charging in step 202.

[0013] In step 204, after a 1-hour cooling period, discharge the battery using a constant current (CC) with a fixed current value of 1C. Once the battery voltage drops from 3.8 volts to 2.8 volts, stop the discharge and let the battery sit and cool for one hour.

[0014] Then in step 206, charge the battery by applying a constant current (CC) of 1C to bring the battery voltage to the higher 4.25 volts. When the battery voltage reaches 4.25 volts, perform a constant voltage (CV) charge, and the current will decrease to keep the battery voltage constant at 4.25 volts. After the charging current drops below a lower threshold, the charging ends, and the battery is allowed to sit and cool for another hour.

[0015] Finally, in step 208, slowly discharge the battery using a constant current (CC) of only 5% of the previous discharge current or 0.05C. This discharge current continues until the battery voltage reaches 2.8 volts. The discharge capacity of the battery is measured by integrating the 0.05C discharge current over the time required to reach the 2.8-volt end point. This integrated current can be compared with the specified charge for a similar test on a new battery to calculate the SOH ratio.

[0016] When the 1C discharge in step 204 exceeds one hour, the 0.05C small current in the discharge step 208 may take a long time, such as 20 hours. The total test time may exceed 26 hours, including the up-to-one-hour rest periods in steps 202, 204, and 206. This long test time is both expensive and undesirable.

[0017] Existing rapid screening methods, such as coulomb counting and internal resistance methods, may be affected by such long test times. The fitting degree of the internal resistance method may be low. These methods may require complex settings.

[0018] There is a need for a screening method for old batteries. It is desired to use a higher current and a smaller voltage range to measure the discharge capacity of old batteries to speed up the test. It is desired to use a combination of constant current and constant voltage methods to more quickly determine the health of the battery. There is a need for a calibration method using Artificial Intelligence (AI) to screen old batteries. Brief Description of the Drawings

[0019] Figure 1 Showing the battery capacity test of the prior art.

[0020] Figures 2A-2B Is the CC-CV charge curve graph of new and old batteries.

[0021] Figures 3A-3B Is the CC-CV charge curve graph using the CCIR method of the parent application and the Sectional-CCIR (SCCIR) method of the present application.

[0022] Figure 4 Is a method for testing and classifying batteries according to the SCCIR ratio measured during the CC-CV charge process.

[0023] Figure 5 Shows the SCCIR test in more detail.

[0024] Figure 6 Shows the process of aging new batteries to obtain SCCIR values for modeling the calibration curve.

[0025] Figure 7 Is a graph showing the functional relationship between the calibration curve of SOH and SCCIR.

[0026] Figure 8 Shows a neural network for modeling the calibration curve of SOH as a function of SCCIR.

[0027] Figure 9 Shows training the neural network with the measured SOH as the target to generate an old battery calibration model.

[0028] Figure 10 Is a calibration sample data table for various combinations of V1, V2, and Imid.

[0029] Figure 11 Is a graph showing the functional relationship between SOH and SCCIR of the calibration sample data for various combinations of V1, V2, and Imid. Detailed implementation manners

[0030] The present invention relates to an improvement in battery screening. The following description is for enabling those of ordinary skill in the art to make and use the present invention in the context of a specific application and its requirements. Various modifications to the preferred embodiments will be apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments. Therefore, the present invention is not intended to be limited to the specific embodiments shown and described, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0031] Figures 2A-2B is the CC-CV charging curve graph of new and old batteries. At Figure 2A , for a new battery, charging is carried out by applying a constant current (CC), such as 0.2C, until the battery voltage, curve 312, reaches V2 at time T1. The constant current can be determined according to the working current or the nominal current defined in the battery specification document. Then at time T1, the charging mode changes from CC to CV. In the constant voltage (CV) mode, the voltage applied to the battery is kept fixed at V2 while adjusting the charging current. In particular, during the CV mode, when the battery is approaching full charge, the charging current, curve 310, decreases from I1. When a certain end point is reached, such as when the charging current drops to a threshold, such as 10% of CC or 0.02C, the CV mode is determined and charging stops.

[0032] At Figure 2B , the same CC-CV method is used to charge the old battery. The old battery stores less electricity than the new battery. Therefore, when the same constant current I1 is applied during the CC mode, Figure 2B the old battery reaches the target voltage V2 at time T1' earlier than Figure 2A the new battery. Figure 2B the voltage curve 312' of the old battery rises to V2 faster than Figure 2A the curve 312 of the new battery.

[0033] The internal resistance of the old battery may increase, and a shorter CC mode time is required. Figure 2B the current curve 310' of the old battery tends to decrease more slowly than Figure 2A the current curve 310 of the new battery.

[0034] The constant current pulse is the initial constant current pulse required for the battery to reach the target voltage. Figure 2A the constant current pulse period of the new battery is time T1, Figure 2B the constant current pulse period of the old battery is time T1'. The charge Qcc supplied to the battery during the constant current pulse is I1×T1 for the new battery and I1×T1' for the old battery.

[0035] The remaining charge Qcv supplied to the battery during CV can be obtained by integrating the current that varies with time during CV. The current curve 310 is integrated from T1 to the end point within the CV phase to obtain Qcv of a new battery, while the current curve 310' is integrated from T1' to the end point within the CV phase to obtain Qcv of an old battery.

[0036] As described in the parent application, the aging or health condition of the battery can be expressed as the constant current pulse ratio (CCIR) of the CC charge to the total charge, that is

[0037] CCIR = Qcc / (Qcc + Qcv).

[0038] The inventors have recognized that the shift of the CC-CV transition point T1 can be used as a measure of battery aging or health. In particular, the inventors measure the CC charge Qcc before T1 and then measure the CV charge Qcv after T1 in order to be able to calculate the constant current pulse ratio (CCIR). Then the CCIR is compared with a calibration curve to determine the state of health (SOH) of the battery. The battery is classified, discarded or reused according to its SOH value.

[0039] Figures 3A-3B is a CC-CV charging curve graph using the CCIR method of the parent application and the local CCIR (SCCIR) method of this application.

[0040] At Figure 3A , as disclosed in the parent application, the measured old battery is first charged with a constant current CC, and the voltage rises from Vmin to Vmax. Once Vmax is reached at time T1', the battery charger switches from the CC mode to the CV mode. The battery voltage remains constant at Vmax, while the battery current drops from the CC value of 0.2C to Imin. Although this CCIR method is useful, the required test time is relatively long because the battery has to be charged from Vmin all the way to Vmax. The inventors wish to reduce the test time.

[0041] At Figure 3B , the local CCIR method charges the old battery within a smaller voltage range, that is, from Vl to V2. Instead of first discharging the old battery to Vmin, the battery is first discharged or charged to a generally higher voltage V1. The target voltage at time T1'' is V2, which is lower than Vmax. The time length of curve 313 is shorter than that of curve 312 because V2 - V1 is less than Vmax - Vmin, making the CC time period shorter. Therefore, Figure 3B the SCCIR method of Figure 3A requires less test time than the CCIR method of

[0042] Rather than charging the battery to a minimum current Imin during the CV mode (as Figure 3A shown), the CV mode is prematurely stopped at a higher target current Imid (as Figure 3B shown). The time length of curve 311 is shorter than that of curve 310 because Imid is closer to the initial current of 0.2C than Imin.

[0043] The inventors have realized that old batteries do not have to be fully discharged and then fully charged. Limited charging within a smaller voltage range can still produce effective results and can be modeled using artificial intelligence (AI). The SCCIR method reduces the test time compared to the CCIR method because the battery is charged within a smaller voltage range and the current change is smaller. The local CCIR method only tests a smaller portion of the full CCIR voltage and current range. The SCCIR method can be used to screen retired batteries more quickly.

[0044] Figure 4 A method for testing and classifying batteries based on the SCCIR ratio measured during the CC-CV charging process. In step 102, the voltage of each old battery is measured as Vcel. In step 104, when Vcel is higher than the maximum voltage Vmax or lower than the minimum voltage Vmin, in step 106, the battery is disposed of.

[0045] In step 104, batteries with an initial voltage Vcel between Vmin and Vmax are further processed. Since the starting voltage V1 is less than Vmax, some batteries may need to be discharged to reach V1. Since V1 is also higher than Vmin, some batteries may need to be charged to reach V1. In step 108, the battery is repeatedly charged or discharged until V1 is reached. A 1-hour rest time can be added after each charge or discharge to allow the battery to cool.

[0046] As Figure 5 shown, an SCCIR test 110 is performed on the old battery to measure the charge Qcc during the CC mode and the charge Qcv during the CV mode. From Qcc and Qcv, the local constant current pulse ratio (SCCIR) is calculated as Qcc / (Qcc + Qcv).

[0047] In step 112, the SCCIR value calculated from the CC and CV charging measurements is compared with a calibration curve to obtain a state of health (SOH) value. A dataset of new batteries is aged ( Figure 6 ) through repeated charge / discharge cycles and an AI model is used to obtain the calibration curve ( Figure 7 ).

[0048] In step 114, the SOH of the battery under test is compared with an SOH threshold (e.g., 80%). In step 106, the batteries with SOH below the threshold are disposed of. In step 116, the batteries with SOH above the threshold are classified into quality grades according to their SOH values. The classified batteries can be reused for various applications based on the quality grades. Some applications may require higher-quality reused batteries than others. For example, compared with the batteries with SOH between 85% and 80%, the batteries with SOH above 95% can fetch a higher price and be used in more demanding applications.

[0049] Figure 5 The SCCIR test is shown in more detail. In step 142, the battery under test is first charged or discharged using a constant current of 1C and / or 0.2C until the target starting voltage V1 is reached. When Vcel is far from V1, a larger 1C current can be used first, and a smaller 0.2C current is used when V1 is reached. Let the battery cool and rest for one hour.

[0050] After the rest period, in step 144, the battery is charged at a constant current (CC) of 0.2C until the target end voltage V2 is reached. The constant current is integrated over time to obtain Qcc. Qcc is stored or otherwise recorded.

[0051] Then the charging switches from the CC mode to the CV mode. The battery voltage is held constant at V2 while the charging current decreases over time to maintain V2. In step 146, once the final charging current Imid is reached, the CV charging mode ends, and the battery rests and cools for 3 minutes. The current drops from 0.2C at the start of the CV phase to Imid at the end of the CV phase, and is integrated over the time of the CV phase to obtain the CV charge Qcv. Qcv is stored in computer memory or otherwise recorded.

[0052] In step 148, the current SCCIR is calculated for the battery under test. The SCCIR is calculated as the ratio of Qcc to Qcc + Qcv. The SCCIR represents the percentage of the total charge in the CC phase. In step 150, the SCCIR value of the battery under test is stored, for example, written into computer memory such as a register file, SRAM, DRAM, or hard disk. Since V2 - V1 is less than Vmax - Vmin, the Qcc of the SCCIR method is less than the Qcc of the CCIR method. Similarly, since Imid is greater than Imin, the Qcv of the SCCIR method is less than the Qcv of the CCIR method.

[0053] Figure 6 The process of aging a new battery to obtain an SCCIR value for modeling a calibration curve is shown. Figure 6The process can be repeated with many new batteries to obtain a data set that can be used to build a model of a calibration curve, which can be used to classify old batteries. Calibrating just 3 batteries provides sufficiently accurate modeling data.

[0054] In step 122, first discharge the new battery to be calibrated at a constant current of 0.2C until the voltage V1 is reached. Let the battery cool and rest for one hour.

[0055] In step 124, after the rest period, charge the battery at a constant current (CC) of 0.2C until the voltage V2 is reached. Integrate the constant current over time to obtain Qcc. Qcc is stored or otherwise recorded.

[0056] Then the charge is switched from the CC mode to the CV mode. The battery voltage is held constant at V2 while the charge current decreases over time to maintain V2. In step 126, once the final charge current Imid is reached, the CV charge mode ends and the battery rests and cools for 3 minutes. Integrate the current from 0.2C at the start of the CV phase down to Imid at the end of the CV phase over the time of the CV phase to obtain the CV charge Qcv. Qcv is stored in the computer memory or otherwise recorded.

[0057] Since V2 - V1 is less than Vmax - Vmin, the battery is only partially charged during the SCCIR test. To obtain an accurate SOH, the full charge capacity of the battery is obtained in steps 136, 140, 128. In step 136, charge the battery at a constant current of 0.2C from V2 to the maximum voltage Vmax. In step 140, continue charging at a constant voltage (CV) of Vmax until the battery current drops to the specified minimum value Imin and the battery is fully charged.

[0058] In step 128, discharge the fully charged battery at a constant current of 0.2C until Vmin is reached. Integrate the 0.2C constant current over time to obtain the current charge capacity Cnow of the aged battery. The current state of health (SOH) of the aged battery is calculated as Cnow / Cinit, where Cinit is the initial charge capacity of the battery before aging, which can be measured in step 128, before any aging occurs in the charge / discharge cycle in step 134.

[0059] Calculate the SCCIR from Qcc obtained in step 124 and Qcv obtained in step 126, i.e., SCCIR = Qcc / (Qcc + Qcv). In step 130, both the SCCIR and the SOH are stored in the computer memory.

[0060] When the SOH is higher than 30% (step 132), the battery is aged by performing 50 charge / discharge cycles using a 1C constant current (step 134). Then the SCCIR sequence is repeated starting from step 122. Through the charge / discharge cycles of step 134, the SOH of the battery gradually decreases. Once the SOH drops below 30% (step 132), the stored SCCIR and SOH data are applied to the AI engine to generate an SOH model as a function of SCCIR, i.e., a calibration curve (step 138).

[0061] Figure 7 is a graph showing the functional relationship between the calibration curve of SOH and SCCIR. Through Figure 6 The SOH and SCCIR data points obtained from the aging process are plotted as points in the graph. The calibration curve 302 is the best-fit function that best fits these data points. Figure 4 In step 112 of, the calibration curve 302 is used to obtain the SOH model value of SCCIR calculated from the measured values during the CC-CV charging of the old battery.

[0062] The calibration curve 302 can be obtained from the AI modeling of these (SOH, SCCIR) data points, such as finding parameters using the least squares method and optimizing using neural networks. Other statistical methods can also be used.

[0063] An artificial neural network (ANN) can be used to generate an SOH model as a function of SCCIR. Artificial neural networks are particularly useful for processing large amounts of non-linear data in complex ways that are difficult to define using traditional computer programs. Instead of being programmed with instructions, training data is input into the neural network and compared with the expected output, then adjusted within the neural network, and the training data is processed and output compared again to generate further adjustments to the neural network. After many such training cycles, the neural network is changed to effectively process data similar to the training data and the expected output. Neural networks are an example of machine learning because the neural network learns how to generate the expected output for the training data. Then real data similar to the training data can be input into the neural network to process real-time data.

[0064] Figure 8Displays a neural network for modeling the calibration curve of SOH as a function of SCCIR. Input node 12 receives the input data SCCIR, while output node 60 outputs the operation result SOH_CALC of the neural network, which is the SOH modeling value of the input SCCIR value. Two layers of operations are performed within this neural network. Nodes 20, 22, 24, …, 28, 29 each receive an input from input node 12, perform a wavelet function operation, and send the output to the nodes in the second layer. The second-layer nodes 52, 54, … 58, 59 also receive multiple inputs, combine these inputs to generate an output, for example by generating a product, and then send the output to the third-layer node 60, which also combines or adds the inputs to generate an output.

[0065] The inputs at each level are typically weighted, so a weighted sum (or other weighted operation result) is generated at each node. A weight can be assigned to each input at a node, which is multiplied by the input, and then all the weighted inputs are added, multiplied, or otherwise operated on by that node to generate the output of that node. These weights for the nodes 20, 22, 24, … 28, 29 in the wavelet layer are designated as A ij 、B ij ,and for the nodes 52, 54, … 58, 59 in the product layer are designated as W ij 。During the training process, these A ij 、B ij 、W ij weight values are adjusted. Through trial and error or other training routines or learning algorithms, ultimately, higher weights can be assigned to the paths that produce the expected output, and lower weights can be assigned to the paths that do not produce the expected output. Machine learning which paths will produce the expected output and assigns high weights to the inputs on these paths.

[0066] These weights can be stored in weight memory 100 or another memory. Since a neural network usually has many nodes, there may be many weights to be stored in weight memory 100. Each weight may require multiple binary bits to represent the range of possible values of that weight. Weights typically require 8 to 16 bits. Weight memory 100 can be SRAM, DRAM, flash memory, a disk, or various combinations of these or other computer storage devices.

[0067] Figure 9 Shows using the measured SOH as a target to train the neural network to generate an old battery calibration model. The aging battery is measured, and at Figure 6In step 130, the measured SCCIR and SOH data are stored. The measured SCCIR data is used as training data 34, SCCIR_MEAS. The measured SOH data corresponding to the SCCIR_MEAS value is recorded as target data 38, SOH_MEAS. Each value of SOH_MEAS corresponds to the value of SCCIR_MEAS, which is measured simultaneously during the process test of aging battery life using Figure 6 The process test of aging battery life

[0068] The neural network 36 receives the training data 34 and the current set of weights A ij 、B ij 、W ij , and operates on the training data 34 to produce a result. The resulting result is the modeled value of SOH, SOH_CALC. The generated result SOH_CALC of the neural network 36 is compared with the target data 38 SOH_MEAS through the loss function 42 to generate a loss value, which is a function of the distance between the generated result and the target. The loss value generated by the loss function 42 is used to adjust the weights applied to the neural network 36. The loss function 42 can apply many iterations of the weights to the training data 34 until the minimum loss value is determined, and the final set of weights for modeling the calibration curve.

[0069] The neural network 36 can have multiple output nodes 60 to generate multiple SOH_CALC values in parallel from the parallel inputs of SCCIR_MEAS, rather than generating a single value of SOH_CALC. The loss function 42 can compare multiple values of SOH_CALC with multiple values of SOH_MEAS in parallel to produce a loss function value.

[0070] Figure 10 is the calibration sample data table for various combinations of V1, V2, and Imid. The SCCIR test may be sensitive to the selection of V1, V2, and Imid. Figure 10 Eight combinations of V1, V2, and Imid are shown in Figure 6 and the battery is calibrated using the calibration method of

[0071] Sample 1 is unacceptable because the SCCIR remains too close to zero as the battery ages. The test time is too short because V2 - V1 is too small for the current used and Qcc is close to 0. Sample 2 is also unacceptable because the SCCIR remains close to 100% and Qcv is close to 0. It is necessary to select V1, V2, and Imid such that the SCCIR is between 0 and 1.0 (0 to 100%) to obtain better test sensitivity, otherwise it is difficult to distinguish batteries close to 80% SOH.

[0072] Figure 11It is a functional relationship diagram of SOH and SCCIR for calibration sample data of various combinations of V1, V2, and Imid. Figure 10 Samples 1 - 8 of Figure 10 are plotted as sample data 701 - 708. When the SOH varies between 70 - 100%, sample 3, data 703, has a nearly vertical line segment at approximately 95% SCCIR. This is undesirable because test noise may create a small change in the measured SCCIR, resulting in a large change in SOH. This is also impractical because it is difficult to distinguish batteries close to the target 80% SOH threshold. The SCCIR values of samples 3 and 4 are close to 1.0 because Qcv drops faster at higher voltages, and reducing the current to 0.19C results in a very small Qcv.

[0073] The test accuracy of sample 3 will be poor. The data 704 of sample 4 also has a nearly vertical linear characteristic, where a slight change in SCCIR within the critical test range will result in an excessive change in SOH. Therefore, samples 3 and 4 are not suitable.

[0074] Compared with the 0.19C current of unacceptable samples 1 - 4, the current of samples 5 - 8 is lower, at 0.15C. The lower current increases the test time but provides better sensitivity. Samples 5 - 8, data 705 - 708, have a flatter slope than samples 3 - 4, data 703, 704. Any one of samples 5 - 8 can be used to set V1, V2, and Imid to generate a calibration curve using Figure 6 the method of

[0075] Sample 5, data 705, has a larger V2 - V1 than samples 6 - 8, data 706 - 708, and a more limited range of SOH and SCCIR values than samples 6 - 8. Nevertheless, any one of samples 5 - 8 can be used to set V1, V2, and Imid. Sample 7, data 707, has the widest SCCIR range, from 40 - 95%, with an SOH of 55 - 100%.

[0076] V2 should be at least 0.2 volts lower than the nominal voltage of the battery, i.e., the average battery voltage during charging, to prevent the SCCIR from dropping linearly to 0. V2 - V1 can be less than 0.25 volts, resulting in a significant reduction in test time compared to the CCIR method. By making Imid at least 0.04C lower than the constant charging current Icc, the SCCIR can be prevented from dropping linearly near 1.

[0077] Alternative Embodiments

[0078] The inventors also envision several other embodiments. For example, the order or sequence of some steps can be changed. As an example, Figure 6In step 130, storing the SCCIR and SOH data can occur during step 128, rather than after that step. There can be various modifications to the neural network, such as having more layers or weights or different functions. More sample points can be input, and more iteration cycles or epochs can be used. Neural network modeling and optimization can be used to obtain a very good fit to the model of the calibration curve 302.

[0079] The calibration curve 302 can be implemented as a look-up table that outputs a modeled SOH value when the measured SCCIR is input to the look-up table. The calibration curve 302 can also be implemented as a function executed by a processor such as a microprocessor, central processing unit, arithmetic logic unit, coprocessor, or other programmed machine. The memory can be shared or separate, local, remote, or various combinations, and the processor and other computing blocks can be shared, distributed, local, remote, or various combinations.

[0080] Although the calibration endpoint as Figure 6 shown in step 132 is based on an SOH threshold, the collection of SCCIR and SOH data can be stopped after a certain number of data points are collected or after a certain time, or when the aging cycle or number of repetitions of step 134 reaches a certain number, or when some other criterion such as Qcc or Qcv becomes 0. For targets above 85%, the calibration curve can collect more than 80% of the SOH. The test technician can simply run out of time and stop further data collection, and then proceed to step 138 to generate a model of the calibration curve 302. An initial model can be generated for use, and then a more refined model can be substituted from more data points.

[0081] Based on SOH, V1, V2, and I Internal , using SCCIR modeling and 0.2C instead of 0.05C, the overall test time for old batteries can be reduced from 26 hours to less than 1 hour. With the improvement in the accuracy of the SOH estimation method, a higher current can be used to make the test time faster.

[0082] Although integrating the current to produce Qcc and Qcv has been described, for the integration of a constant current, the constant current can be multiplied by the time period during which the constant current is applied. Various approximate integration methods can be applied, such as using Piece-Wise-Linear (PWL) or multiplying the current by each of several short time periods. Coulomb counting methods can be used to integrate the charge over time. The integration method can accumulate the charge transferred over a small time period.

[0083] Although an initial deep discharge is not required, if needed, the battery can be pre-discharged or pre-charged in other steps. The battery can initially be discharged or charged to Vmin instead of the usually higher voltage V1. The rest time can be shortened or extended. A simple battery bench test setup can be used instead of a complex test bench. SCCIR can be defined not as Qcc / (Qcc + Qcv), and alternatively, SCCIR can be defined as Qcv / (Qcc + Qcv), and the calibration curve 302 can be adjusted according to the new definition.

[0084] The calibration curve can be approximated by one or more functions, such as piecewise linear (PWL) or multivariate functions. SOH can be modeled by equations with terms such as the square root, logarithm, etc. of SCCIR.

[0085] The temperature of the battery during testing should be maintained at a constant value, such as room temperature. The length of the rest time after battery charging or discharging can depend on the charge / discharge current and the thermal performance of the battery. The thermal characteristics of the battery may change with age, for example, the heat generation of old batteries increases due to increased internal resistance.

[0086] Many parameters and values can be changed from the given examples. Voltages such as Vmax, Vmin, Vnominal, V2, etc. and currents C, I1 can have different values or different ratios to each other. For Figure 10 the samples of V1, V2, Imid shown, Imin can be 0.02C, Vmax can be 4.2 volts, Vmin can be 2.75 volts, Vnominal can be 3.7 volts, and this is just one of many examples. The difference between V2 - V1 can be a very small value, such as 0.25V or less. V2 can be lower than the battery voltage at 80% State-Of-Charge (SOC), higher than the nominal voltage by -0.2 volts, V2 > V1. The CV charge current endpoint Imid can be between Icc - 0.04C and 0.1Icc, where Icc is the CC charge current used.

[0087] The number of discharge / charge cycles used in each step during the aging process can be adjusted to other values, such as 10 cycles, 100 cycles, etc., depending on the required accuracy of the calibration curve 302. The charge current (Qcc + Qcv) during screening between V1 and V2 can be less than one-quarter of the current charge capacity Cnow measured during calibration, allowing for faster screening because the batteries being screened are only partially charged, charged one-quarter. Qcc + Qcv can be even smaller, such as one-tenth of Cnow, depending on the accuracy of the testing machine. V2 - V1 can be one-quarter or one-tenth or less of Vfull - Vempty (i.e., the battery fully charged - empty voltage).

[0088] The number of batteries for test calibration can be a relatively small number, such as 3 batteries when the AI modeling is effective, or more batteries, such as 100 batteries, can be tested when using a less effective modeling or when more precise calibration is required. Some battery reuse applications may not require precise SOH modeling. Ideally, the batteries for test calibration match well with the batteries to be screened, such as the same manufacturer and model. The batteries under test can be single batteries or battery packs, single cells or multiple cells.

[0089] Depending on the application or intended use of the reused battery, some test errors can be tolerated. In some cases, a test error of + / - 3% of the actual SOH may occur. When a larger current can be used to achieve the required test accuracy or error tolerance, the test time may be shortened.

[0090] Some embodiments may not use all components. Other components can be added. The loss function 42 can use various error / loss and cost generators, such as a weight decay term that prevents the weights from growing too large over multiple cycles of training optimization; a sparsity penalty, on the other hand, can encourage nodes to set their weights to zero, so that only a small fraction of the total nodes are like this. Many alternatives, combinations, and variations are possible. Other variations and loss or cost terms can be added to the loss function 42. The relative scale factor values of different cost functions can be adjusted to balance the effects of various functions. The training end point of the neural network can be set for various combinations of conditions, such as the desired final accuracy, accuracy - hardware cost, target hardware cost, etc.

[0091] Various combinations of software, hardware, firmware, routines, modules, functions, etc. can be used to implement the neural network 36, the loss function 42, and other components in a variety of technologies. The final result, the calibration curve 302, or the calibration function generator can be derived from the neural network 36 with the final weights and can be implemented as a program module or in a dedicated integrated circuit (ASIC) or other hardware to improve the processing speed and reduce power consumption.

[0092] The background section of the present invention may contain background information about the problems or environment of the present invention, rather than describing the prior art of others. Therefore, the materials included in the background art section are not an admission by the applicant of the prior art.

[0093] Any method or process described herein is machine-implemented or computer-implemented and is performed by a machine, computer, or other device, rather than solely by a human without machine assistance. Tangible results produced can include reports or other machine-generated displays on display devices such as computer monitors, projection devices, audio generation devices, and associated media devices, and can include hardcopy printouts that are also machine-generated. Computer control of other machines is another tangible result.

[0094] Any advantages and benefits described may not apply to all embodiments of the present invention. When the term "means" appears in a claim element, the applicant intends that the claim element fall within the provisions of 35 USC Section 112, Paragraph 6. Typically, there is a label of one or more words before the term "means". The one or more words before the term "means" are a label for the purpose of facilitating reference to the claim element and not for expressing a structural limitation. Such means-plus-function claims are intended to cover not only the structures described herein for performing the function and their structural equivalents, but also equivalent structures. For example, although nails and screws have different configurations, they are equivalent structures because they both perform the fastening function. Claims that do not use the term "means" do not fall within the provisions of 35 USC Section 112, Paragraph 6. A signal is typically an electrical signal, but can also be an optical signal, for example, that can be transmitted over a fiber optic line.

[0095] The foregoing description of the embodiments of the present invention has been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many modifications and variations are possible in light of the above teaching. The intention is that the scope of the invention not be limited by this detailed description, but rather by the limitations appended to the claims.

Claims

1. A method for screening batteries for reuse or disposal, comprising: (a) Charging the battery at a constant current from a starting voltage until the battery voltage reaches a voltage threshold; wherein the voltage threshold is less than the full voltage of the battery when fully charged; wherein the starting voltage is between the full voltage of the battery when fully charged and the empty voltage when the battery is fully discharged, or the starting voltage is equal to the empty voltage; (b) Multiplying the value of the constant current by a constant current CC time period during which the constant current is applied to the battery to reach the voltage threshold to produce a constant current charge value Qcc, and storing the constant current charge value in a computer memory; (c) After reaching the voltage threshold, partially charging the battery at a constant voltage until a variable current applied to the battery voltage reaches a current threshold greater than a minimum current Imin; (d) Integrating the value of the variable current over a constant voltage CV time period during which the constant voltage and the variable current are applied to the battery to reach the current threshold to produce a constant voltage charge value Qcv, and storing the constant voltage charge value in the computer memory; (e) Generating a local constant current pulse ratio SCCIR by dividing the constant current charge value Qcc by the sum of the constant current charge value Qcc and the constant voltage charge value Qcv, i.e., SCCIR = Qcc / (Qcc + Qcv); Inputting the local constant current pulse ratio SCCIR into a calibration function processor, and the calibration function processor outputs a model state of health SOH value corresponding to the local constant current pulse ratio SCCIR input to the calibration function processor, and the model state of health SOH value is a function of the local constant current pulse ratio SCCIR; When the model state of health SOH value is higher than a state of health SOH threshold, classifying the battery using the model state of health SOH value for reuse; when the model state of health SOH value is lower than the state of health SOH threshold, classifying the battery for disposal.

2. The method according to claim 1, wherein a first difference between the voltage threshold and the starting voltage is less than one quarter of a second difference between the full voltage and the empty voltage; Among them, Accelerating battery screening to generate the local constant current pulse ratio SCCIR by partially charging the battery only below one quarter of the full voltage range of the battery.

3. The method according to claim 2, wherein Step (a) further includes: Discharging the battery to the starting voltage before applying the constant current, and charging the battery from the starting voltage to the voltage threshold during the constant current CC time period.

4. The method according to claim 3, further comprising: Before discharging the battery to the activation voltage, the battery is pre-screened by measuring an initial voltage of the battery, and when the initial voltage of the battery is less than a minimum pre-screening voltage, the battery is discarded; wherein the constant current does not exceed 20% of the maximum battery current.

5. The method according to claim 3, further comprising: in response to the model state of health SOH value, classifying the battery into one of a plurality of bins, wherein each of the plurality of bins receives batteries having different ranges of model state of health SOH values.

6. The method according to claim 3, further comprising: generating an SCCIR-SOH model that programs the calibration function processor to generate the model state of health SOH value from an input local constant current impulse ratio SCCIR value by: (f) performing steps (a) to (e) on a new battery and storing the local constant current impulse ratio SCCIR as a local constant current impulse ratio SCCIR value of a model input; (g) fully charging the new battery to the full voltage and then discharging the new battery at the constant current during a capacity measurement period that ends when the new battery reaches a second voltage threshold; (h) multiplying the value of the constant current by the capacity measurement period to produce a current capacity charge value and storing it in the computer memory; (i) generating a model input state of health SOH value by dividing the current capacity charge value by an initial capacity charge value, wherein the initial capacity charge value is the current capacity charge value before the new battery is aged by step (j); (j) aging the new battery by repeatedly charging and discharging the new battery in N charge / discharge cycles, where N is an integer of at least 10; repeating steps (f) to (i) after the new battery is aged; using the model input local constant current impulse ratio SCCIR value and the model input state of health SOH value to generate parameters describing the SCCIR-SOH model, the SCCIR-SOH model programming the calibration function processor to generate the model state of health SOH value from an input local constant current impulse ratio SCCIR value.

7. The method according to claim 6, wherein using the model input local constant current impulse ratio SCCIR value and the model input state of health SOH value to generate parameters describing the SCCIR-SOH model further comprises: (m) inputting the model input local constant current impulse ratio SCCIR value into an input end of a neural network; processing the model input local constant current impulse ratio SCCIR value using the neural network to produce a calculated state of health SOH value; comparing the calculated state of health SOH value with the model input state of health SOH value using a loss function to produce a loss value; Use the loss value to adjust the weights of the nodes within the neural network and repeat from step (m) until a modeling end point is reached; Store the weights in a computer memory connected to the neural network; When the modeling end point is reached, use the final value of the weights and the neural network to generate the model state of health SOH value from the input of the local constant current pulse ratio SCCIR, so as to implement the calibration function processor for generating the model state of health SOH value from the input local constant current pulse ratio SCCIR value.

8. The method according to claim 7, wherein the neural network includes a first layer of nodes that execute a wavelet function, a second layer of nodes that execute a product function, and a third layer of nodes that execute a summation function.

9. The method according to claim 7, further comprising: (k) Compare the state of health SOH value of the model input with the end state of health SOH value; When the state of health SOH value of the model input is greater than the end state of health SOH value, continue from step (j) to continue aging the new battery; When the state of health SOH value of the model input is equal to or less than the end state of health SOH value, jump to step (m) to generate the SCCIR-SOH model.

10. A battery screening method, comprising: First, discharge the battery using an initial constant current until a first voltage target is reached; Cool the battery for a period of time after the initial discharge; Partially charge the battery using a constant current until a second voltage target is reached, and record a constant current charge Qcc, that is, the amount of charge transferred to the battery through the constant current when the battery goes from the first voltage target to the second voltage target; When the second voltage target is reached, switch to a constant voltage CV charging process and record a constant voltage charge Qcv, that is, the amount of charge transferred to the battery during the constant voltage CV charging process; Calculate the local constant current pulse ratio SCCIR, that is, the ratio of Qcc to the sum of Qcc and Qcv, that is, SCCIR = Qcc / (Qcc + Qcv), or the ratio of Qcv to the sum of Qcc and Qcv, that is, SCCIR = Qcv / (Qcc + Qcv); input the local constant current pulse ratio SCCIR into a calibration function generator, and the calibration function generator returns a model state of health SOH value corresponding to the input local constant current pulse ratio SCCIR value; Use the model state of health SOH value to determine when to discard the battery and when to reuse the battery; Accordingly, the battery is screened according to the model state of health SOH value, and the model state of health SOH value is a function of the measured local constant current pulse ratio SCCIR.

11. The battery screening method according to claim 10, wherein using the constant current to partially charge the battery until the second voltage target is reached further includes: Transfer a partial charge to the battery, and the partial charge is less than one-fourth of the full charge of the battery; Wherein the battery screening time is reduced by only partially charging the battery to generate the local constant current pulse ratio SCCIR.

12. The battery screening method according to claim 10 further includes: During the constant voltage CV charging process, allowing the charging current to vary while maintaining a constant voltage applied to the battery; And When the charging current reaches a mid - current target, terminating the constant voltage CV charging process.

13. The battery screening method according to claim 12 further includes: Generating calibration data points by aging and measuring a plurality of new batteries, each new battery being processed by a calibration data collection process, the calibration data collection process including: (a) Initially charging or discharging the new battery using an initial constant current until the first voltage target is reached; Cooling the new battery for a period of time after the initial discharge; Partially charging the new battery using the constant current until the second voltage target is reached, and recording a constant current charge Qcc, i.e., the amount of charge transferred to the new battery through the constant current; When the second voltage target is reached, switching to the constant voltage CV charging process and recording a constant voltage charge Qcv, i.e., the amount of charge transferred to the new battery during the constant voltage CV charging process; Calculating a data - point local constant - current pulse ratio SCCIR, i.e., the ratio of Qcc to the sum of Qcc and Qcv, i.e., SCCIR = Qcc / (Qcc + Qcv), or the ratio of Qcv to the sum of Qcc and Qcv, i.e., SCCIR = Qcv / (Qcc + Qcv); Fully charging the new battery to a full voltage, the full voltage being greater than the first voltage target; Fully discharging the new battery using the constant current until an empty voltage is reached, where the empty voltage is less than the second voltage target, and recording the current charge capacity, i.e., the integral of the constant current discharged from the new battery; Generating a data - point state of health SOH value, i.e., the ratio of the current charge capacity to the initial charge capacity of the new battery before aging; Storing the data - point local constant - current pulse ratio SCCIR and the data - point state of health SOH value as a calibration data point; Aging the new battery by charging and discharging the new battery N times, where N is an integer of at least 10; Repeating from step (a).

14. The battery screening method according to claim 13 further includes: When the data - point state of health SOH value is less than a threshold state of health SOH value, ending the calibration data collection process.

15. The battery screening method according to claim 13, wherein, The current charge capacity is at least ten times the sum of Qcc and Qcv, where Qcc is measured between the first voltage target and the second voltage target, and the current charge capacity is measured between the full voltage and the empty voltage; Wherein a first difference between the first voltage target and the second voltage target is less than one - quarter of a second difference between the full voltage and the empty voltage; Wherein by only partially charging the battery to generate the local constant - current pulse ratio SCCIR, the battery screening time is reduced.

16. The battery screening method according to claim 15 further includes: Using multiple calibration data points as inputs to a model generator that programs the calibration function generator; (b) Inputting multiple data point local constant current pulse ratios SCCIR as inputs into an input layer of a neural network, the neural network generating multiple computed outputs based on the inputs and multiple weights; Using a loss function to compare the multiple computed outputs with multiple data point state of health SOH values to adjust the multiple weights; Using the adjusted values of the multiple weights, repeating from step (b) until the loss function reaches an end point; When the end point is reached, applying the multiple weights to the neural network to generate a computed output for an input of a local constant current pulse ratio SCCIR, the computed output being the model state of health SOH value of the calibration function generator.

17. A method for estimating the state of health SOH of a battery, comprising: Discharging / charging the battery using an initial constant current until the battery reaches a first voltage target; Charging the battery using a constant current until the battery reaches a second voltage target and determining a Qcc charge, i.e., the amount of charge transferred to the battery through the constant current; After reaching the second voltage target, continuing to charge the battery using a variable current to maintain a constant voltage across the battery; Determining a Qcv charge, i.e., the amount of charge transferred to the battery through the variable current; Calculating a local constant current pulse ratio SCCIR, i.e., the ratio of the Qcc charge to the sum of the Qcc charge and the Qcv charge, i.e., SCCIR = Qcc / (Qcc + Qcv); Inputting the local constant current pulse ratio SCCIR into a processor that outputs a model state of health SOH, the model state of health SOH being a function of the local constant current pulse ratio SCCIR; Comparing the model state of health SOH with a state of health SOH threshold as a basis for disposing of or reusing the battery by classification; Accordingly, classifying the battery based on the model state of health SOH determined from the measured local constant current pulse ratio SCCIR of the battery.

18. The method for estimating the state of health SOH of a battery according to claim 17, further comprising: Aging a new battery through multiple charge / discharge cycles, collecting data points, and measuring its local constant current pulse ratio SCCIR values and state of health SOH values; Generating a calibration curve that best fits the data points; The processor using the calibration curve to generate the model state of health SOH from the local constant current pulse ratio SCCIR input to the processor.

19. The method for estimating the state of health SOH of a battery according to claim 18, further comprising: Discharging the new battery at the constant current until a low voltage target is reached, the low voltage target being less than the first voltage target, and integrating the constant current to produce a current charge capacity; The state of health (SOH) value of the new battery that has undergone the aging, which is the ratio of the current charge capacity to the initial current charge capacity before aging; where (Qcc + Qcv) is less than one quarter of the current charge capacity, where during the estimation period, the battery is only partially charged between the first voltage target and the second voltage target, and during the calibration period, it is fully charged when generating the current charge capacity.

20. The method for estimating the state of health (SOH) of a battery according to claim 19 further includes: Input the data points into a neural network to generate the calibration curve.

Images