Estimation of battery health and charge level

Abstract

A system and method for estimating the state of health (SOH) and / or state of charge (SOC) of a battery that does not require time-consuming charge / discharge cycles. The test system uses a machine learning-trained estimation model to provide an estimate of the battery's SOH or SOC based on battery test results for specific battery characteristics, based on battery chemistry. When battery chemistry is input, the test system automatically applies tests to the battery and calculates values ​​for specific characteristics based on the test results. These values ​​for specific characteristics are then used by the trained estimation model to estimate the battery's SOH or SOC. The system is based on a data-driven approach to estimating the battery's state of health and / or SOC based on the selection of specific characteristics (or parameters) and the association of values ​​for these characteristics with the battery's state of health and / or SOC.

Description

[Technical Field]

[0001] [Cross-reference of related applications] This application claims priority to U.S. Provisional Patent Application No. 63 / 490,448, filed on 15 March 2023, entitled “BATTERY STATE OF HEALTH ESTIMATION,” the entirety of which is incorporated herein by reference.

[0002] [Technical field] This invention relates to the health and charge state of a battery. More specifically, this invention relates to a system and method for estimating the health or charge state of a battery without requiring time-consuming charge / discharge cycles or time-consuming measurements of full charge / discharge capacity. [Background technology]

[0003] While the number of electric vehicles (EVs) and plug-in electric vehicles (PHEVs) worldwide is expected to increase significantly, it is necessary to investigate the potential uses of used EV batteries, especially those with a remaining capacity exceeding 70-80% of their original capacity, particularly given their high cost. EV batteries, after an average of 10 years of use or warranty replacement, are recyclable for material recovery, but other more economical options exist. One option is to remanufacture end-of-life (EOL) EV batteries and reuse them in mobile applications. A second option is to reuse end-of-life EV batteries for stationary energy storage applications such as backup power for residential and commercial buildings, peak shaving, arbitrage, and balancing intermittent renewable energy generation.

[0004] Refurbished batteries are available in vehicles to reduce vehicle battery costs after preliminary screening and diagnosis of the battery pack modules and cells. This refurbishment process involves partial disassembly of the battery pack or module, removal of substandard cells, replacement of these cells, and reassembly of the module and pack. In addition to the growth of the EV and PHEV markets, advances in the lifespan of EV batteries and the potential benefits of energy storage integrated into the power grid are creating new room for the secondary use of EV batteries in stationary applications (in addition to the primary use of EV batteries in vehicle-to-X applications, where X refers to homes, buildings, or the grid).

[0005] The design and implementation of prognosis and health management (PHM) architectures are considered key to addressing battery endurance issues. Deploying PHM solutions enables failure prediction and avoidance, assessment of state of health (SOH), and estimation of remaining service life (RUL), allowing for the development of control and / or maintenance measures to ensure the continuity of battery operations based on this information. However, different battery states such as state of charge (SOC), SOH, and RUL cannot be directly observed. These typically require estimation and predictive algorithms that provide battery diagnostics and forecasts.

[0006] The primary role of diagnostics is to estimate failure parameters to determine the system's End-of-Life (EOL) once a failure mode is detected. For electric vehicle applications, this limit is generally set at 80% of the nominal capacity. Other studies define a different limit based on internal resistance when the internal resistance increases to 160% of its initial value, under the same SOC and operating temperature conditions. Predictive models, on the other hand, enable EOL prediction based on failure parameter estimation and SOH diagnostics of the battery storage system. For lithium-ion batteries, this predictive model should be able to obtain real-time information on aging parameters such as capacity and internal resistance in order to estimate the SOH of the battery storage system online. The primary role of the predictive model is to predict future degradation based on the diagnostic results and to present the remaining service life of the resource at a particular operating point by estimating the battery's RUL (Return on Life).

[0007] To determine whether EOL batteries should be recycled or remanufactured / reused, it is necessary to diagnose, screen, and select EV batteries using safe, reliable, and efficient methodologies.

[0008] Current State of Health (SOH) assessment techniques rely on determining the battery's health by operating it through multiple charge / discharge cycles. Unfortunately, such tests can take many hours to reach a proper SOH. In addition, rapid charge / discharge cycles can significantly degrade the battery's condition.

[0009] Regarding the state of charge (SOC) of a battery, it is known that the SOC is the ratio of the capacity (ampere-hour) available in the battery at a specific time to the total capacity of the battery. In international air transportation, the SOC is limited to 30% of the rated capacity of the battery. The transportation guidelines for Li batteries specify that Li-ion batteries and batteries must not be used for transportation in a state of charge (SOC) exceeding 30% of their rated capacity. However, prior to transportation by aircraft, railway, etc., there is no portable SOC estimator that can be used to quickly and reliably check the battery SOC at an airport or at a railway station somewhere. Therefore, there is a need for a portable tool that provides a quick and reliable, off-board prediction of the SOC of the battery being transported.

[0010] Therefore, there is a need for a system and method that provides the health state of the battery, the state of health (SOH) of the battery, and the SOC of the battery faster than current technologies while still providing useful results. SUMMARY OF THE INVENTION

[0011] The present invention provides a system and method for estimating the state of health (SOH) and / or state of charge (SOC) of a battery without requiring time-consuming charge-discharge cycles or full charge-discharge cycles. A test system uses a machine-learned estimation model to provide an estimated value regarding the SOH of the battery or the SOC of the battery based on the results of battery tests regarding specific battery characteristics based on battery chemistry. The test system automatically applies a test to the battery when the battery chemistry is input and calculates a value regarding a specific characteristic based on the test result. The value regarding the specific characteristic is then used by the learned estimation model to estimate the SOH of the battery or the SOC of the battery. The system is based on a data-driven approach for estimating the state of health and / or state of charge of a battery based on the selection of specific characteristics (or parameters) and the association of values regarding these characteristics with the state of health and / or state of charge of the battery.

[0012] In a first aspect, the present invention provides a method for estimating the state characteristics of a battery, the method comprising: a) receiving an input regarding the components of the battery; b) determining, based on the input received in step a), the conditions to be used in a test of the battery; c) configuring an automated test device connected to the battery and for testing the battery, based on the conditions, such that the test device performs a test of the battery regarding specific characteristics using the conditions; d) automatically testing the battery using the test device to obtain test results; e) deriving a specific characteristic of the battery from the test results obtained in step d); e) using a learned model to estimate the state characteristics of the battery based on the specific characteristic, wherein the state characteristics of the battery are Battery health and, Battery charge status It is at least one of the following.

[0013] In a second aspect, the present invention provides a system for testing or estimating the state characteristics of a battery, the system is A database having entries for different battery chemistrys, and for specific characteristics of each battery chemistry used to estimate the above state characteristics of a battery for each of those different battery chemistrys, and also including entries for test conditions used when testing each battery chemistry, A test module for testing a battery, wherein the test applied by the test module is based on the above-mentioned specific characteristics relating to the battery chemistry of the battery and the above-mentioned test conditions relating to the battery chemistry of the battery, A trained model for estimating the above state characteristics of a battery based on calculated specific characteristics, A main processor for receiving battery-related inputs, the main processor being connected to the database, the test module, and the trained model, and the inputs being used to determine the battery chemistry of the battery, In this system, The main processor retrieves specific characteristics and test conditions from the database regarding the battery chemistry of the battery. The main processor above configures the test module to apply appropriate tests to the battery based on the specific characteristics and test conditions extracted with respect to the battery chemistry of the battery above. The above test module sends the test results of the appropriate test to the above main processor, and the main processor calculates the above-calculated specific characteristics from the test results. The specific characteristics calculated above are sent to the trained module by the main processor. The above state characteristics of the battery are, Battery health, and Battery charge status It is at least one of the following. [Brief explanation of the drawing]

[0014] Embodiments of the present invention will now be described with reference to the following drawings, where the same reference numerals in different drawings indicate the same elements.

[0015] [Figure 1] This is a block diagram of a system according to one aspect of the present invention. [Figure 2] This figure shows the current-voltage profile of a pulsed current test used to test a battery, according to another aspect of the present invention. [Figure 3] This is a flowchart detailing the steps of a method according to another aspect of the present invention. [Figure 4A] This figure shows various resistance values ​​measured as a function of SOC, illustrating their use in SOH estimation. [Figure 4B] This figure shows various resistance values ​​measured as a function of voltage V0. [Figure 5] This flowchart details the steps involved in constructing and using one embodiment of a SOH / SOC estimation database, as well as its use during battery testing. [Modes for carrying out the invention]

[0016] It is clear that known onboard SOC estimations can utilize data collected during operation of electric vehicles (EVs) and stationary battery energy storage systems (BESS). However, offboard SOC estimation may not always require the use of battery management system (BMS) information and data collected from battery operation. Offboard SOC estimation can utilize simple test protocols, fast estimation algorithms, and portable tools to leverage limited datasets collected indoors or outdoors.

[0017] Referring to Figure 1, a block diagram of a system according to one aspect of the present invention is shown. The system 10 includes a main processor 20, a database 30, a test module 40, and a trained estimation module 50. The test module 40 is coupled to a battery 60 and is capable of testing the battery and collecting test results from the test.

[0018] The main processor 20 receives the chemistry and other parameters of the battery under test. The processor 20 then uses the specific battery chemistry and other parameters from the database to determine which specific characteristics should be tested / derived from the battery 60, and which conditions and / or parameters should be used when testing the battery 60, based on the battery chemistry.

[0019] Once the conditions and characteristics for the test are determined by the processor, these conditions and characteristics are used to configure the test equipment of the test module 30 so that the appropriate test (and test parameters) can be automatically applied to the battery 60.

[0020] The test module 30 applies the test to the battery 60, and the test results are sent to the processor 20. The processor 20 then derives specific characteristics from these test results. The calculated values ​​for various specific characteristics are then sent to the trained estimation module 50. The module 50 then determines which battery SOH / battery SOC corresponds to these calculated values ​​and outputs the thus estimated battery SOH / estimated battery SOC 70.

[0021] The system for SOH estimation is not based on calculations of healthy state or charge / discharge cycles, but rather on a data-driven approach. Rather than calculating values ​​based on test results (such as battery capacity assessed through repeated charge / discharge cycles) and determining this as a percentage of rated capacity, the correlation between experimental data collected regarding the healthy state of the battery and specific characteristics of the battery under test is determined. As is well known, current techniques calculate the ratio of sensed battery capacity to rated capacity to determine the healthy state. This invention is based on the analysis of experimental data collected for batteries of a specific chemistry to determine which specific characteristics indicate a pseudo-linear correlation with the healthy state of the battery. These specific characteristics are then calculated / tested for the specific battery. The calculated / tested values ​​obtained from the tests are then passed through a trained model to estimate the healthy state of the specific battery. For clarification, the trained model is trained using machine learning techniques on a library of experimentally determined data, including specific characteristics and their corresponding SOH value indications. Each experimentally determined battery dataset, along with the battery chemistry, provides further insight into other relevant battery conditions (such as manufacturer and model year) for each battery dataset used to train the estimation module 50. In this way, each battery dataset tracks numerous relevant characteristics of the battery over a wide voltage range as the battery is tested, providing a final SOH assessment for each dataset. These conditions and characteristics are useful for determining relevant test parameters and tests for a given battery chemistry, and the relationship between those parameters and the battery's SOH.

[0022] Similarly, the system for SOC estimation is based on a data-driven approach, rather than on calculating the charge state or the state of charge / discharge cycles to measure full charge / discharge capacity. Instead of calculating values ​​based on test results (such as the full charge / discharge capacity of a battery assessed through charge / discharge cycles, etc.) and instead of determining the usable capacity level (ampere-hours) of a battery relative to its total capacity, the correlation between experimental data collected regarding the battery's charge state and specific characteristics of the battery under test is determined. As is well known, current techniques calculate the ratio of sensed usable battery capacity to total capacity to understand the charge state. This invention is based on the analysis of experimental data collected for batteries of a specific chemistry to determine which specific characteristics show a pseudo-linear correlation with the battery's charge state. These specific characteristics are then calculated / tested for a specific battery. The calculated / tested values ​​obtained from the tests are then passed through a trained model to estimate the charge state of a particular battery. For clarity, the trained model is trained using machine learning techniques on a library of experimentally determined data containing indicated values ​​for specific characteristics and their corresponding SOC values. Each experimentally determined battery dataset provides further insights into other relevant battery conditions (such as manufacturer and model year) along with the battery chemistry for each battery dataset used to train the estimation module 50. In this way, each battery dataset tracks numerous relevant characteristics of the battery over a wide voltage range as the battery is tested, providing a final SOC assessment for each of the datasets. These conditions and characteristics are useful for determining relevant test parameters and tests for a given battery chemistry and the relationship between those parameters and the battery's SOC.

[0023] Referring to Figures 4A and 4B (for use in SOH estimation), the battery dataset shows that the individual characteristics under consideration are fairly constant across a wide range of conditions throughout the test. The example shown in Figure 4A displays various resistance values ​​at a given SOC value of the battery. In the example in Figure 4A, such information shows a profile with significantly less consistency in the range of approximately 5% to 25% SOC. This allows for a better analysis of those values ​​within this “distinguishable zone” of SOC value. Similarly, Figure 4B shows a similar distinguishable zone across a range of voltage V0 levels. If the observed values ​​change linearly across a more clearly distinguishable zone, these values ​​within that distinguishable zone can more directly correlate with SOH, allowing for a simpler correlation. The trained estimation model gives unique weights to the information about the characteristics under consideration for the test that occurs within the above distinguishable zones when determining the relationship between those values ​​and the maximum SOH. By limiting the target voltage in this way, subsequent battery tests can be performed over a much narrower voltage range than in classic full SOH estimation tests. In the example shown in Figure 4A, SOC is used in the test, and in the example shown in Figure 4B, voltage V0 is used, but other voltage values ​​and ranges can be selected based on similar "identifiable zone" determinations where the target data displays a profile that is more easily mapped to SOH. As mentioned above, Figures 4A and 4B relate to SOH estimation.

[0024] The trained model learns to estimate the SOH / SOC for a given battery, given specific values ​​for these particular characteristics within a discernible zone for a given battery chemistry.

[0025] Regarding the example shown in Figure 4B, for clarification, preferably, several voltages or at least one voltage within a specific voltage range to obtain effective parameters for LMO / NMC LIBs during the discharge process are selected, such as 3.5~3.7V (preferred V: 3.65 / 3.60 / 3.55V). Similarly, for different chemistry, the preferred voltage range during discharge is: LMO / NMC: 3.50~3.70V, LFP: 3.0~3.15V, NMC111: 3.25~3.50V, NMC622: 3.25~3.45V, NCA: 3.0~3.30V That is the case.

[0026] To clarify further, tests applied to a battery may include pulsed current tests, in which a pulsed current flows through the battery and the voltage obtained at a specific point in time after the current has flowed is sensed. Specific battery characteristics may include specific resistance and capacitance, which can be calculated based on the voltage and current sensed from the pulsed current test. As mentioned above, different battery chemistry may require different specific characteristics. Therefore, different battery chemistry may have different specific characteristics that correlate with the battery's SOH / SOC.

[0027] Furthermore, it is clear that optimal test conditions can differ for different battery chemistrys. Therefore, for some battery chemistrys, optimal test results may be obtained when the results fall within a specific, identifiable zone. These test conditions and specific characteristics used in SOH / SOC assessment are all battery chemistry-dependent and stored in a database. Thus, when the system receives the chemistry of the battery under test, the relevant specific characteristics and test conditions used for the battery are retrieved from the database. These test conditions and specific characteristics are then used to construct a test module, if necessary. For example, if one or more specific characteristics of a battery require multiple voltage readings at multiple current inputs, the number of times the current passes through the battery, the levels of those currents, and the voltage readings are used to construct the test module. Similarly, if it is known that voltage readings within a specific, identifiable zone yield the best results for a particular battery chemistry, such a desired range is used for data collection to calculate the parameters used to construct the test module. Thus, different current levels may be applied to the battery under test until voltage readings within the desired range are read. These voltage readings, along with other data points sensed for the battery, are subsequently stored as test results for that battery. Other data points may include current, voltage, temperature, and how these readings develop / change over time while the battery is being tested.

[0028] In one embodiment of the present invention, the test applied to the battery may take the form of a pulsed current test. This test involves applying a current pulse to the battery using a pulse having a specific fixed duration. At a specific point in time after the pulse is applied, the battery voltage reading is obtained. The pulse is applied at an initial level that is not zero. This means that when a current pulse is applied, a non-zero current is supplied to the battery. The current and voltage profiles for this current pulse are shown in Figure 2.

[0029] As can be seen in Figure 2, a current pulse at a level of delta I1, higher than the initial non-zero current level, is applied for 30 seconds and then cut off at the zero current level. Voltage readings are taken at 0, 2, 10, 18, 30, 32, 40, 48, 60, and 90 seconds after the current pulse is applied. The current pulse level may be read or known. The voltage and current readings can be used to calculate the associated resistance / impedance values ​​as needed. This current and voltage profile for a current pulse test is just one example of a profile that may be applied to a battery. In some runs, the current pulse and associated readings are taken multiple times, and the current pulse levels may differ for different pulses.

[0030] As you can see, each application of the pulsed current test takes only 90 seconds. In one run, the test is repeated three times to obtain a dataset. As you can imagine, the time required for three 90-second tests is considerably less than the several hours required for charge-discharge cycles that current battery testing techniques require.

[0031] In a single run, SOH estimation may be sufficient using a single current injection for 30 seconds at a specific voltage during the charging or discharging process. A set of current injection and current interruption is good for comparing parameters in two different modes, although the current interruption mode is arbitrary. In a single run, with respect to the test time frame, the test requires only 30 seconds to a maximum of 4.5 minutes of current injection at one selected voltage as the pulse current test time (i.e., 90 seconds x 3 for three different voltages). Note that this time frame does not include the preconditioning period to reach the selected test voltage during charging (to a unidirectional EV charger) or during discharging (for all common test equipment or conditions, including bidirectional EV charger applications).

[0032] Regarding pre-adjustment, pre-adjustment may involve the following steps: 1) a voltage pre-adjustment step, where the obtained voltage is increased or decreased to a selected voltage range for a particular battery chemistry at a selected initial current; and 2) after reaching the selected voltage, a current injection step, where a pulsed current is applied for 30 seconds at a specific current based on the selected C rate (current level), followed by a current cutoff step, where the pulsed current is cut off for 30-60 seconds for data acquisition. Therefore, if the battery voltage is not within the specified voltage range for a particular chemistry, the battery must be charged or discharged to this specific voltage range before applying the pulsed current. If the initially measured battery voltage is too high to apply the pulsed current for SOH estimation, the battery may be discharged, and the selected current may be applied to result in a voltage drop.

[0033] Regarding specific characteristics used to estimate the state of health (SOH) of a battery, analytical and experimental data indicate that the following specific characteristics may be used in estimating the SOH / SOC of a battery. R o - Ohmic Resistor R CT -Charge transfer resistance e R p - Polarization resistance R IN - Total internal resistance CPE l -Double layer capacity V0 - The last voltage before current injection due to an initial current that is not zero.

[0034] The six parameters mentioned above can be calculated from test data or measured directly from test data, but a seventh parameter, a specific characteristic or function of a parameter, may also be used. Based on the analysis of the training battery dataset, certain challenges have been identified. While general battery chemistry is nearly constant for each battery, there may be some variability in recorded characteristics between different model years or different manufacturers. The specific causes of these slight variability are unknown, but it is thought that this may be due to small changes in the sourcing of chemicals or components, or other similar differences resulting from changes in processes over time at a given manufacturer. Due to these slight variability, it becomes more difficult to properly formulate a more general relationship between battery chemistry and SOH / SOC estimation, and these slight variability can accumulate and lead to unacceptable errors in SOH / SOC estimation over time. Therefore, if the battery under test directly matches the chemistry and battery state of a sufficient number of training battery datasets, the above six parameters are sufficient to estimate the State of Health (SOH). However, for a broader and more comprehensive estimation, additional functions are needed.

[0035] If the battery chemistry is the same but the battery conditions are different among the training data sets, the seventh parameter is determined by referring to a number of training battery data sets. In collecting this training battery data set, the six identified parameters can be compared with each other to find a relationship function that shows the best pseudo-linearity across the entire subset of the training data set. This function can then be used to define a seventh parameter that functions in a similar way to a smoothing function to reduce errors across different battery conditions, thereby providing a more generalized function for estimating the SOH for a given battery chemistry.

[0036] Such a function can be determined using high-quality experimental data and machine learning / AI techniques. Similarly, simple determination methods can be used to determine specific characteristics and functions of specific characteristics. Such a method was used in one implementation and useful results comparable to more time-consuming curve fitting methods were obtained.

[0037] Note that the use of the seventh parameter is extremely effective in enhancing the reliability and accuracy of SOH estimation. The seventh parameter having the best pseudo-linearity as a function of the SOH value can be identified by comparing the linearity of the combinations of the first to sixth parameters and using appropriate constants in a specific formula as shown below. However, note that this seventh parameter has different formulas for different battery chemistries due to different charging and discharging behaviors. The seventh parameter is simply a function of the ohmic resistance, charge transfer resistance, and total internal resistance f(R o +R ct ,R total ) where R o +R ct = ohmic resistance + charge transfer resistance, <00002​​​​​​​​Furthermore, specific characteristics may be determined using a resistor-capacitor equivalent circuit modeling suitable for the battery.

[0039] It is clear that, in order to generate the specific characteristics and parameters to be used for each battery chemistry, numerous analyses of SOH data and specific characteristic data are necessary for each battery chemistry. Similarly, for batteries with specific chemistry manufactured by a particular manufacturer, a specific set of characteristics or parameters unique to them may also be required for use in SOH testing and estimation.

[0040] When a pulsed current test is applied to the battery under test as needed, and when a desired range of voltage / indicated values ​​is achieved, the test results are obtained. As described above, pulsed current tests are used at different current levels to obtain indicated voltages within the desired range. Once the indicated voltages within the desired range are achieved, the current levels that produce that desired voltage range are used by repeating the application of the pulsed current test. The indicated voltages for these repeated applications, as well as the indicated current, time, and temperature, form a test result dataset for the battery under test. This test dataset can then be used to calculate values ​​used in estimating the battery's state of health (SOH).

[0041] Regarding SOC estimation, the pulse current profiles and selected parameters (except for the seventh parameter) described above are all the same as those used for SOH estimation, but it is clear that they are used in the full charge / discharge voltage state of a battery with a specific chemistry. Therefore, for SOC estimation, the range of full voltages for charging / discharging can be used for pulse current testing. This voltage range will, of course, differ for different battery chemistrys. Regarding the seventh parameter for SOC estimation, this seventh parameter is simply a function of the last voltage before current injection with a non-zero initial current, and the voltage selected for a specific battery chemistry. f(V0,V C ) And here, V0 = the last voltage before current injection due to an initial current that is not zero. V c =Voltage selected for a specific battery chemistry That is the case.

[0042] In some embodiments, the method and system may also include means for checking the temperature and / or current (or C rate) applied to the battery under test during the pulsed current test as a preliminary condition for the test. Each of these factors is one of the first five parameters (R O , R CT , R P , R IN CPE l This may affect the measurement of SOH or SOC tests (f(R) O +R ct ,R total ) and f(V0,V c This affects the determination of the respective seventh parameter for either of the following: )) The model may then include separate lookup tables for normalizing or correcting the above parameters for battery temperature and applied current (or C rate) to be used when generating the estimated SOH or SOC values.

[0043] As shown in Figure 5, a trained model that estimates battery SOH / SOC using calculated values ​​can be trained using a BRANN (Bayesian Regularized Artificial Neural Network) and a dataset suitable for a battery of a particular chemistry, which has a dataset containing values ​​for specific characteristics and their corresponding battery SOH / SOC values. Of course, other neural network / machine learning techniques for training a suitable model can also be used with such datasets. A trained model learns to provide SOH / SOC estimates based on the training dataset when given a set of values ​​for a specific characteristic. As you can imagine, the quality of the training dataset can be very important for the SOH / SOC estimates produced.

[0044] In one run, a dataset of LMO / NMC EV batteries was analyzed, and the six to seven specific characteristics mentioned above were selected for SOH estimation. For batteries with such chemistry (LMO is lithium-ion manganese battery, NMC is nickel-manganese-cobalt battery), experiments and data show that readings in the range of 3.5 to 3.7V (preferably 3.65 / 3.60 / 3.55V) yield test results showing a significant reduction in the time required for SOH estimation. Such runs produce SOH estimates that are within 1-2% error of the measured SOH (as determined using the charge / discharge method).

[0045] It should be noted that the above range of 3.5 to 3.7V is clearly for the SOH estimation during the discharge process. If the initial battery voltage is higher than 3.7V, the selected I will discharge the battery to 3.5 to 3.7V as part of the pre-adjustment process. O (For example, 0.5C) applies. That is, OThis means there is a certain correlation between the initial voltage and the selected test voltage range. In the implementation of this invention, during the discharge process, a 1.0C rate is applied for the current injection step at 3.5–3.7V. For initial voltages lower than 3.5V, a charging step may be applied to reach a specific voltage range. Due to the overall overpotential of the battery in the charging direction, the selected voltage range may shift to a higher voltage range during the charging process. A selected voltage range of 3.5–3.7V is effective for estimating the State of Health (SOH) during the battery discharge process. The overpotential of the battery in the charging direction depends on the battery's chemistry and structure. Therefore, a higher voltage range (e.g., 3.65–3.80V) may be used for pulse current testing during the charging process of LMO / NMC batteries.

[0046] When estimating the State of Charge (SOC) of a battery under test, it is not necessary to use a specific voltage range of the selected battery. Pulse current testing can be performed within the battery's full charge / discharge voltage range for SOC estimation.

[0047] Furthermore, it should be noted that the above invention also allows a typical battery cycler to be used in conjunction with appropriate automation components in a test module. Thus, the test module may use a combination of a power supply, voltmeter / multimeter, electronic load for pulsed current, and configurable automation components so that a desired voltage range can be automatically achieved while applying pulsed current testing. Once the desired voltage range is achieved, appropriate repeated applications of the pulsed current test can be automatically applied to obtain a useful test results dataset. In the above implementation, once the desired voltage range was achieved, a useful test results dataset was collected within 5 minutes.

[0048] In terms of execution, the main processor can run using a properly programmed data processing unit, such as a microcontroller, ASIC, or any suitable general-purpose data processing unit. The database can be a suitable data storage device that operates to store a suitable amount of data and is connectable to the main processor. The trained model can be a software module, a properly programmed ASIC, or a dedicated processor configured to operate according to the logic of the trained model, as determined by training.

[0049] Referring to Figure 3, a flowchart detailing the steps of a method according to one aspect of the present invention is shown. As can be seen, the method is performed by the system shown in Figure 1. The method begins with step 100, which receives the battery chemistry of the battery under test (via the main processor). Step 110 is the step of determining the specific characteristics and test conditions to be used when testing the battery of a given battery chemistry. This is done by retrieving the relevant entries from a database. Once the specific characteristics and test conditions are retrieved, the test module is then configured to apply the necessary tests and perform the test conditions (step 120). As can be imagined, this includes settings such as the specific voltage range to be investigated when testing the battery, the length of time the pulse current is applied, and when the voltage value is acquired. It is obvious that these and other test parameters and conditions may differ for different battery chemistrys (and for different battery manufacturers).

[0050] After the test module is properly configured, the battery under test may be pre-conditioned so that it is properly configured / conditioned for the required / desired test (step 130). As described above, the pre-conditioning portion of the test (which may be performed before the test is applied or included in the test step) may include charging the battery, discharging the battery, or ensuring that the battery is in a suitable charge state for the test. Step 140 is the application of the test to the battery. Once the desired range of the indicated value (i.e., the indicated voltage) is achieved, the test is repeated (step 150) with the values ​​necessary to achieve that range of indicated value. This results in a (possibly multiple) test result dataset.

[0051] Once a dataset of test results (which may be multiple) is generated, these datasets are then sent from the test module to the main processor, where values ​​for specific characteristics are calculated as needed (step 160). For example, relevant resistance, impedance, and capacitance are calculated using the indicated values ​​of voltage, current, temperature, time, and other data from the dataset. The calculated values ​​for specific characteristics are then sent to a trained estimation module, which generates the estimated SOH / SOC (step 170).

[0052] Compared to other methods for estimating battery SOH / SOC, this method and system enables significantly faster SOH / SOC estimation while producing equally accurate results. In validation comparisons with other techniques, the results of this method consistently exhibit a mean squared error (MSE) of less than 2% for the measured SOH / SOC value across multiple chemistrys of the battery under test. Thus, this method and system achieves highly accurate SOH / SOC estimation in a much shorter time.

[0053] Certain embodiments of the present invention may be used as part of a larger battery-centric system. For example, the SOH / SOC diagnostic process of the present invention may be incorporated into a battery monitoring and control system, such as an energy management system (EMS) within a battery management system (BMS) and / or battery energy storage system. Similarly, the process of the present invention may be implemented in a centralized computer for online-based SOH / SOC estimation and in a unidirectional / bidirectional electric vehicle supply system (EVSE) for local SOH / SOC estimation of EV batteries.

[0054] For clarification, in order to perform SOH estimation, the estimation method begins by receiving the battery chemistry of the battery under test. Then, specific characteristics and test conditions to be used when testing the battery of a given battery chemistry are determined. This is done by retrieving relevant entries from a database. Once the specific characteristics and test conditions are retrieved, the test module is then configured to apply the necessary tests and perform the test conditions. This step includes setting specific voltage ranges to be investigated during battery testing, the length of time pulse current is applied, when voltage values ​​are acquired, etc. It is clear that these and other test parameters and conditions may differ for different battery chemistrys (and also for different battery manufacturers). As an example, some voltage ranges for testing a particular battery chemistry are as follows: LMO / NMC: 3.50~3.70V LFP: 3.0~3.15V NMC111: 3.25~3.50V NMC622: 3.25~3.45V NCA: 3.0~3.30V

[0055] After the test module is properly configured, the battery under test may be pre-conditioned so that it is suitablely configured / conditioned for a given / desired test. This pre-conditioning portion of the test (which may be performed before the test is applied or included in the test step) may include charging the battery, discharging the battery, or ensuring that the battery is in a suitable charge state for the test (based on battery chemistry). This pre-conditioning portion may also include checking the battery temperature. After the battery has been properly pre-conditioned, the test is applied to the battery. This test may be a current pulse test.

[0056] For one run of this current pulse test, a current pulse is applied for 30 seconds at a non-zero current level and then cut off at zero current. Voltage values ​​are taken at 0, 2, 10, 18, 30, 32, 40, 48, 60, and 90 seconds after the current pulse is applied. The current pulse level may be read or known. The voltage and current readings may be used to calculate the associated resistance / impedance values ​​as needed. The current and voltage profiles for the current pulse test are only one possible profile that may be applied to the battery. The current pulse and its associated readings may be taken multiple times, and the current pulse levels may differ for different pulses. Part of the current pulse test may include determining / checking the current flowing between various repetitions / applications of the current pulse test.

[0057] After the desired range of the indicated value (i.e., the voltage indicated value) is measured and recorded, the test is repeated with the values ​​necessary to achieve this range of indicated value and thereby produce a test result dataset (which may be multiple).

[0058] The dataset (which may be multiple) is then used to calculate values ​​for specific characteristics as needed. For example, the indicated values ​​for voltage, current, temperature, time, and other data from the dataset are used to calculate the associated resistance, impedance, and capacitance. These specific characteristics may include the following six: R o - Ohmic Resistor R CT - Charge transfer resistance R p - Polarization resistance R IN - Total internal resistance CPE l -Double layer capacity V0 - The last voltage before current injection due to a non-zero initial current.

[0059] The seventh characteristic is a function of ohmic resistance, charge transfer resistance, and total internal resistance. f(R o +R ct ,R total ) And here, R o +R Ct = Ohmic resistance + Charge transfer resistance, R total = Total internal resistance That is the case.

[0060] The values ​​of these specific characteristics may need to be normalized or corrected during current pulse testing based on the sensed battery temperature and / or the applied current. As described above, each of these elements is one of the first five parameters (R o , R CT , R P , R IN , and CPE l This may affect the measurement of the SOH test (f(R o +R Ct ,R totalThis may influence the determination of each of the seventh parameters related to ). Then, one or more lookup tables for elements that normalize or correct the specific characteristics described above (based on sensed battery temperature and based on applied current (or C rate)) are referenced, and suitable values ​​for such normalization or correction elements are determined. These correction / normalization element values ​​are applied to the indicated values ​​for the specific characteristics as needed.

[0061] These various (after being corrected or normalized as necessary) calculated values ​​for specific characteristics are then sent to a trained estimation module containing a trained model that estimates the battery SOH. Training may be performed using a BRANN (Bayesian Regularized Artificial Neural Network) and a dataset suitable for a battery of a particular chemistry, which has a dataset containing values ​​for specific characteristics and their corresponding battery SOH values. The trained model learns to provide an SOH estimate based on the training dataset, given a set of values ​​for specific characteristics.

[0062] For clarification, in order to perform SOC estimation, the estimation method begins by receiving the battery chemistry of the battery under test. Next, specific characteristics and test conditions to be used when testing a battery of a given battery chemistry are determined. This is done by retrieving the relevant entries from a database. Then, using the retrieved specific characteristics and test conditions, the test module is configured to apply the necessary tests and to execute the test conditions. This step includes setting specific voltage ranges to be investigated when testing the battery, the length of time pulse current is applied, when voltage values ​​are acquired, etc. It should be noted that these and other test parameters and conditions may differ for different battery chemistrys (and for different battery manufacturers). For clarification, the tests for SOC estimation are performed over the entire full charge-discharge voltage range, and this charge-discharge voltage range may differ for different battery chemistrys. Thus, for battery chemistry A, the battery may be considered to be discharged at voltage ν1 and fully charged at voltage ν2. Thus, with respect to this battery chemistry A, the test can be performed at any voltage between ν1 and ν2.

[0063] After the test module is properly configured, the battery under test may be pre-conditioned so that it is properly configured / tuned for the required / desired test. This pre-conditioning portion of the test (which may be performed before the test is applied or included in the test step) may include charging the battery, discharging the battery, or verifying that the battery is in a suitable charge state for the test (based on battery chemistry) (i.e., that the battery charge is within the charge-discharge voltage range for that battery chemistry). As part of the pre-conditioning, this may include checking the battery temperature. After the battery has been properly pre-conditioned (if necessary), the test is applied to the battery. This test may be a current pulse test.

[0064] For one run of this current pulse test, a current pulse is applied for 30 seconds at a non-zero current level, and then cut off at a zero current level. After the current pulse is applied, voltage values ​​are taken at 0, 2, 10, 18, 30, 32, 40, 48, 60, and 90 seconds. The current pulse level may be read or known. The voltage and current readings may be used to calculate the associated resistance / impedance values ​​as needed. Note that this current and voltage profile for the current pulse test is only one possible profile that may be applied to a battery. Multiple current pulses and associated readings may be taken, and the current pulse levels may differ for different pulses. Part of this current pulse test may include determining / checking the current flowing during various iterations / applications of the current pulse test.

[0065] After the desired range of the indicated value (i.e., the voltage indicated value) is measured and recorded, the test is repeated with the values ​​necessary to achieve that range of indicated value and thereby produce a test result dataset (which may be multiple).

[0066] The (potentially multiple) datasets are then used to calculate values ​​for specific characteristics, as needed. For example, the indicated values ​​for voltage, current, temperature, time, and other data from the dataset are used to calculate the associated resistance, impedance, and capacitance. These specific characteristics may include the following six: R o - Ohmic Resistor R CT - Charge transfer resistance R p - Polarization resistance R IN - Total internal resistance CPE l -Double layer capacity V0 - The last voltage before current injection due to a non-zero initial current.

[0067] The seventh characteristic of SOC estimation is a function of the last voltage before current injection with a non-zero initial current and the voltage selected for a particular battery chemistry. f(V0,V C ) And here, The last voltage before current injection with an initial current where V0 is not zero. Vc = Voltage selected for a specific battery chemistry That is the case.

[0068] The values ​​of these specific characteristics may need to be normalized or corrected during current pulse testing based on the sensed battery temperature and / or the applied current. As described above, each of these elements is one of the first five parameters (R o , R CT , R P , R IN , and CPE l This may affect the measurement of the SOC test (f(V0,V C This may influence the determination of each of the seventh parameters related to ). Then, one or more lookup tables for elements that normalize or correct the specific characteristics described above (based on sensed battery temperature and based on applied current (or C rate)) are referenced, and suitable values ​​for such normalization / correction elements are determined. These correction / normalization element values ​​are applied to the indicated values ​​for the specific characteristics as needed.

[0069] These various (after being corrected or normalized as necessary) calculated values ​​for specific characteristics are then sent to a trained estimation module containing a trained model that estimates the battery SOC. Training may be performed using a BRANN (Bayesian Regularized Artificial Neural Network) and a dataset appropriate for a battery of a particular chemistry, containing a dataset with values ​​for specific characteristics and their corresponding battery SOC values. The trained model learns to provide SOC estimates based on the training dataset, given a set of values ​​for specific characteristics.

[0070] It is clear that various aspects of the present invention can be executed as software modules within an entire software system. Therefore, when executed, the present invention may take the form of computer executable instructions that execute various software modules having predefined functions.

[0071] Embodiments of the present invention may be carried out by a computer processor or a similar device programmed as a method step, or by an electronic system equipped with means for carrying out these steps. Similarly, electronic storage means such as computer diskettes, CD-ROMs, random access memory (RAM), read-only memory (ROM), or similar computer software storage media known in the art may be programmed to carry out such method steps. Furthermore, electronic signals representing these method steps may be transmitted over a communication network.

[0072] Embodiments of the present invention are executable in any conventional computer programming language. For example, preferred embodiments are executable in a procedural programming language (e.g., "C" or "Go") or an object-oriented language (e.g., "C++", "Java", "PHP", "Python", "C#"). Alternative embodiments of the present invention are executable as pre-programmed hardware elements, other related components, or combinations of hardware and software components.

[0073] Embodiments are executable as computer program products for use with computer systems. Such execution may include a set of computer instructions fixed to either a tangible medium such as a computer-readable medium (e.g., a diskette, CD-ROM, ROM, or hard disk) or a medium that can be transmitted to a computer system via a modem or other interface device such as a communication adapter connected to a network via a medium. The medium may be either a tangible medium (e.g., an optical communication line or a telecommunication line) or a medium executed using wireless technology (e.g., microwave, infrared, or other transmission technology). The set of computer instructions embodies all or part of the functions described herein. Those skilled in the art will understand that such computer instructions can be written in numerous programming languages ​​for use with many computer architectures or operating systems. Furthermore, such instructions can be stored in any memory device, such as semiconductor, magnetic, optical, or other memory devices, and can be transmitted using any communication technology, such as optical, infrared, microwave, or other transmission technology. Such computer program products are envisioned to be distributed as removable media (e.g., shrink-wrapped software) with attached printed or electronic documentation, pre-loaded onto a computer system (e.g., system ROM or fixed disk), or distributed from a server over a network (e.g., the Internet or the World Wide Web). Of course, some embodiments of the present invention are executable as a combination of both software (e.g., computer program products) and hardware. Yet another embodiment of the present invention is executable entirely as hardware or entirely as software (e.g., computer program products).

[0074] Those who understand the present invention can now conceive of alternative structures and embodiments or variations to all of the above, all of which shall fall within the scope of the present invention as defined in the appended claims.

Claims

1. A method for estimating the state characteristics of a battery, wherein the method is a) Receiving input regarding the components of the battery, b) Based on the input received in step a), determine the conditions to be used in testing the battery, c) Configuring an automated test device connected to the battery and for testing the battery, such that the test device uses the conditions to test the battery with respect to specific characteristics, d) To obtain test results, the battery is automatically tested using the test equipment, e) Deriving the specific characteristics of the battery from the test results obtained in step d), f) Using a trained model to estimate the state characteristics of the battery based on the specific characteristics, Includes, The state characteristics of the aforementioned battery are: The healthy state of the aforementioned battery, The charge state of the aforementioned battery A method that is at least one of the following.

2. The input received in step a) relates to the chemical components of the battery, as described in claim 1.

3. The method according to claim 1, wherein the conditions include a voltage range for testing the battery.

4. The method according to claim 1, wherein the specific characteristics of the battery include at least one of ohmic resistance, charge transfer resistance, polarization resistance, total internal resistance, double-layer capacitance, the last voltage before current injection with a non-zero initial current, and a function of the specific characteristics or parameters.

5. The method according to claim 1, wherein the test equipment uses a pulsed current method to test the battery.

6. The pulsed current method according to claim 5, wherein the pulsed current method applies a DC current pulse with a non-zero initial current, and obtains a voltage value that forms part of the test result at a specific point in time after the current pulse has been applied.

7. The method according to claim 6, wherein the current pulse is sustained for 30 seconds, and the pulsed current method includes passing an applied current and, after the applied current has been initially passed, interrupting the applied current for 30 seconds.

8. The method according to claim 6, wherein a voltage value is acquired at a specific time after the current pulse is applied, the specific time includes 0 seconds, 2 seconds, 10 seconds, 18 seconds, 30 seconds, 32 seconds, 40 seconds, 48 ​​seconds, 60 seconds, and 90 seconds.

9. The method according to claim 5, wherein step d) comprises repeatedly applying the pulse current method within a specific period of time.

10. The pulse current method according to claim 5, comprising obtaining current and voltage indications for the battery, which form part of the test results, at a specific point in time after the applied current has flowed through the battery.

11. The method according to claim 1, wherein the trained model is trained using a machine learning method.

12. The method according to claim 1, wherein the specific characteristics are selected using at least one of a simplified determination method and a register-capacitor equivalent circuit model.

13. The method according to claim 1, wherein the trained model correlates the state characteristics of the battery with the values ​​relating to the specific characteristics by training the trained model using machine learning techniques for values ​​relating to the specific characteristics.

14. The method according to claim 1, further comprising step e) determining at least one of the temperature of the battery during testing and the current applied (or C rate), wherein the temperature or current applied is for use in determining the particular characteristic.

15. A system for testing or estimating the state characteristics of a battery, wherein the system is A database having entries indicating multiple entries for different battery chemistrys and for specific characteristics of each battery chemistry used to estimate the state characteristics of the battery for each of the different battery chemistrys, and also including entries for test conditions used when testing each battery chemistry, A test module for testing a battery, wherein the test applied by the test module is based on the specific characteristics relating to the battery chemistry of the battery and the test conditions relating to the battery chemistry of the battery, A trained model for estimating the state characteristics of the battery based on calculated specific characteristics, A main processor for receiving input related to the battery, wherein the main processor is connected to the database, the test module, and the trained model, and the input is used to determine the battery chemistry of the battery, The main processor retrieves specific characteristics and test conditions from the database with respect to the battery chemistry of the battery. The main processor configures the test module to apply appropriate tests to the battery based on specific characteristics and test conditions extracted with respect to the battery chemistry of the battery. The test module sends the test results of the appropriate test to the main processor, and the main processor calculates the calculated specific characteristics from the test results. The calculated specific characteristics are sent by the main processor to the learned module. The state characteristics of the aforementioned battery are: The healthy state of the aforementioned battery, and The charge state of the aforementioned battery A system that is at least one of the following.

16. The system according to claim 15, wherein the test module includes at least one of a power supply, a voltmeter, a multimeter, a component for applying an electronic load to the battery, and an automation component for automating the appropriate test.

17. The system according to claim 15, wherein the test includes a pulsed current test.

18. The system according to claim 17, wherein the test conditions include a voltage range to be achieved for the pulse current test.

19. The system according to claim 15, wherein the trained model is trained using machine learning.

20. The system according to claim 15, wherein the specific characteristics include at least one of ohmic resistance, charge transfer resistance, polarization resistance, total internal resistance, double-layer capacitance, the last voltage before current injection due to a non-zero initial current, and a function of the specific characteristics or parameters.

21. The system according to claim 17, wherein the pulsed current test includes applying a current pulse with a non-zero initial current.

22. The system according to claim 17, wherein the pulse current test involves applying a DC current pulse with a non-zero initial current and acquiring a voltage value at a specific point in time after the current pulse has been applied.

23. The method according to claim 1, wherein between steps c) and d), the battery is pre-adjusted so that it is suitable for the test.

24. The system according to claim 15, wherein the battery is pre-adjusted so that it is suitable for the test.

25. The system according to claim 15, wherein at least one of the temperature of the battery during testing and the current (or C rate) applied is used to determine the specific characteristics.

Images