System and method for battery power state estimation using impedance measurements
Through electrochemical impedance spectroscopy measurement and supermodel training, the problem of low battery maximum power state estimation efficiency in the battery management system is solved, and fast and accurate power state prediction and model calibration are achieved, which improves the efficiency and reliability of the battery management system. .
Patent Information
- Application Number
- CN202380036762.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-29
- Filing Date
- 2023-04-28
- Publication Date
- 2024-12-06
AI Technical Summary
Existing battery management systems have inefficiencies in estimating the maximum power state of the battery pack, traditional methods take a long time and require frequent calibration of equivalent circuit models after the battery ages, resulting in impracticality.
Using electrochemical impedance spectroscopy (EIS) measurement technology, supermodel is trained by multiple scans of the battery in different states. It is used to predict the power state of the battery and perform multiple scans before and after the battery is used to calibrate the model to improve the estimation. accuracy and efficiency.
It realizes rapid and accurate estimation of the power state of the battery, reduces the calibration frequency of the equivalent circuit model after battery aging, and improves the efficiency and reliability of the battery management system.
Smart Images

Figure CN119096152A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application is based on a non-provisional utility patent application of U.S. Provisional Application Serial No. 63 / 336,724 filed on April 29, 2022, entitled “System and Method for Battery State Power Estimation Using Impedance Measurement,” the entire contents of which are incorporated herein by reference. Technical Field
[0003] The present invention relates generally to battery monitoring and, more particularly, to a system and method for estimating a battery state of power (SoP) using battery impedance measurements.
[0004] With the rapid development of electric vehicles, battery management systems (BMS) are expected to track the state of the battery pack to enable effective operation of the corresponding control algorithms. One state that needs to be monitored is the maximum power that can be extracted from the battery pack (during rapid acceleration) or put back into the battery pack (during regenerative braking), which directly translates to vehicle performance and user experience. Currently, BMS uses simple and conservative estimates calculated as the product of the minimum terminal voltage and the maximum current in a given time. However, this greatly underestimates the maximum power metric.
[0005] The State of Power (SoP) quantifies the maximum amount of charge a battery can deliver in a short period of time. Traditional battery management systems infer information like SoP by training models based on time-domain voltage and current data, typically in the form of a schedule of current pulses (i.e., direct current internal resistance (DCIR) measurements), where SoP estimation is achieved through long DCIR measurement tests involving current pulses with long relaxation times. This can be very time consuming and impractical for several reasons, one of which is that the resulting equivalent circuit model (ECM) must be recalibrated after the battery has aged slightly. Therefore, there is a need for improved SoP measurements and improved ECM recalibration. Summary of the invention
[0006] A simplified summary of one or more aspects is presented below in order to provide a basic understanding of these aspects. This summary is not an extensive overview of all contemplated aspects, and is neither intended to identify key or important elements of all aspects, nor to describe the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to a more detailed description presented later.
[0007] In this exposition, we define a super model as a set of models, each of which characterizes the voltage response of a corresponding battery of a plurality of batteries of a vehicle battery to a current curve at different battery states. Therefore, for any given battery state, we have an equivalent state space model that accurately represents the electrical characteristics of the battery. According to one aspect, a method for pre-training a super model is provided, which can be used to predict the power state (SoP) of a vehicle battery. The method includes performing electrochemical impedance spectroscopy (EIS) scans on a plurality of batteries similar to the vehicle battery at various states of the battery. In one aspect, the various battery states can include at least some of different temperature ranges, different states of charge (SoC), battery life, and properties of the current load to which the battery is subjected. The method also includes fitting the parameters of the super model by applying optimization techniques to the results of the EIS scan.
[0008] According to another aspect, a method for predicting the state of power (SoP) of a battery is provided. The method includes performing a plurality of electrochemical impedance spectroscopy (EIS) scans on the battery before the battery is used for the first time in a vehicle. The method also includes calibrating a pre-trained super model using the results of the EIS scans. The pre-trained super model includes a family of models, each model defining a voltage response of a corresponding cell among a plurality of cells of the battery to a current distribution under various states of the battery. The method also includes performing a plurality of additional EIS scans on the battery after the battery is used for the first time in the vehicle. The method also includes recalibrating the pre-trained super model using the results of the additional EIS scans.
[0009] According to another aspect, a system for predicting the state of power (SoP) of a battery is provided. The system includes an electrochemical impedance spectroscopy (EIS) system for performing multiple EIS scans on a battery before the battery is first used in a vehicle, and multiple additional EIS scans on the battery after the battery is first used in a vehicle. The system also includes a storage device for storing program code. The system also includes a processing device operably coupled to the EIS system and the memory device, for running program code to calibrate a pre-trained super model using the results of the multiple EIS scans. The pre-trained super model includes a family of models, each model defining a voltage response of a corresponding battery cell among a plurality of battery cells of the battery to a current distribution under various states of the battery. The processing device also runs program code to recalibrate the pre-trained super model using the results of the multiple additional EIS scans.
[0010] To accomplish the above and related ends, one or more aspects include the features hereinafter fully described and particularly pointed out in the claims. The following description and the accompanying drawings set forth in detail certain illustrative features of one or more aspects. However, these features are indicative of only a few of the various ways in which the principles of the various aspects may be employed, and this description is intended to include all such aspects and their equivalents. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] The disclosed aspects will be described below in conjunction with the accompanying drawings, which are used to illustrate rather than limit the disclosed aspects, wherein like reference numerals represent like elements, dashed lines may represent optional elements, and wherein:
[0012] Figure 1 is a schematic diagram of a battery system within a vehicle according to an exemplary aspect;
[0013] Figure 2 is a schematic diagram illustrating a battery equivalent circuit model (ECM) according to an exemplary aspect;
[0014] Figure 3 is a diagram showing an example Nyquist plot of an EIS imaginary part versus a real part of impedance according to an exemplary aspect, showing the measured impedance of a battery and the Figure 2 ECM estimated impedance;
[0015] Figure 4 is a diagram showing an exemplary aspect of Figure 2 A graph of the terminal voltage of the battery ECM versus time;
[0016] Figure 5 is a schematic diagram illustrating another battery ECM according to an exemplary aspect;
[0017] Figure 6-8 10-11 are diagrams showing the calculation of the state of charge (SoC) and temperature (T) using the nonlinear least squares method (NLLS) according to an exemplary aspect. Figure 5 Plot of the estimated parameters of the ECM in ;
[0018] Fig. 9 It is a Nyquist plot showing the negative imaginary part of the battery impedance and the real part of the battery impedance;
[0019] Fig.12 is a graph of open circuit voltage (OCV) versus SoC according to an exemplary aspect;
[0020] Fig.13 is a diagram illustrating extended Kalman filter (EKF) functionality according to an exemplary aspect;
[0021] Fig.14 is a block diagram illustrating a shutdown use case according to an exemplary aspect;
[0022] Fig.15 is a block diagram illustrating key usage according to an exemplary aspect;
[0023] Fig.16 Shown Figure 2 The current and time relationship diagram of the ECM of the middle cell;
[0024] Fig.17 Shown Figure 2 A graph showing the relationship between the voltage and time of the ECM of the middle battery;
[0025] Fig.18 is a flow chart illustrating a method for offline pre-training of super model parameters according to an exemplary aspect;
[0026] Fig.19 is a flow chart illustrating a method for calibrating and recalibrating a new battery according to an exemplary aspect;
[0027] Figure 20-21 It is further shown that according to the exemplary aspect Fig.19 Flow chart of the steps of the method. DETAILED DESCRIPTION
[0028] In various aspects, systems and methods are provided for estimating the state of power (SoP) of a battery using electrochemical impedance spectroscopy (EIS) measurement techniques. In various aspects, impedance measurements taken at multiple frequencies are used to estimate the battery state of power (SoP). Battery SoP is defined as the maximum power that a battery can produce or absorb at any point in time without exceeding manufacturer limitations, such as maximum current, maximum and minimum terminal voltages, and temperature. In various aspects, the impedance measurements can be used to train a family of equivalent circuit models (ECMs), labeled "ECM super models." The super model defines the voltage response of each battery to its respective current profile under different conditions (e.g., different combinations of temperature and state of charge, or in another example, different combinations of battery age). While the cell is operating, the model is used to track the internal state of the cell, accessing the SoP estimate of the cell at any point in time through a prediction mechanism. Alternatively, a static model can be defined for relevant inputs measured in the field.
[0029] Thus, the hypermodel may be an equivalent circuit model (ECM) that maps the various resistive and capacitive components of the ECM to different combinations of temperature and state of charge used in an EIS scan. While one or more aspects may involve modeling a battery using a battery equivalent circuit model (ECM), other representations may be used, such as, but not limited to, a model simplified from a physics-based model that maps physical representations of battery elements to various combinations of temperature and state of charge used in an EIS scan, etc. It should be understood that aspects of the present disclosure may use any type of battery model to obtain an SoP estimate in accordance with the present disclosure.
[0030] Aspects described herein are intended to report how much power a battery can produce or absorb at any point in time using information from electrochemical impedance spectroscopy (EIS), as exceeding certain limits on power applied to an electric vehicle battery during charge / discharge can lead to serious problems such as premature aging, electrochemical failure, or thermal runaway. Additionally, conventional BMS systems infer SoP-like information by training a model based on time-domain voltage and current data, typically in the form of a current pulse time table (i.e., direct current internal resistance (DCIR) measurement). In another example, a conventional BMS system infers SoP-like information by training a model based on time-domain voltage and current data from a hybrid pulse power characterization (HPPC) ammeter. Additionally, EIS scan measurements can more completely and accurately simulate the behavior of a battery by training a model on an impedance spectrum.
[0031] EIS is an emerging technique in battery management systems (BMS). In EIS, cells are stimulated with a sinusoidal current and the resulting voltage is measured (or vice versa). This is repeated at many different frequencies to produce an impedance spectrum that can be used to fit a model that characterizes the electrochemical performance of the battery.
[0032] Aspects described herein involve collecting impedance spectra while the car is at rest (“off”) at various state of charge (SoC) levels and temperatures, and using this data to train a family of equivalent circuit models (ECMs), referred to herein as “ECM super-models.” Impedance spectra measurements may also be collected on the battery before it is placed in the car, and the model calibrated using field measurements (“off”). The super-model defines the voltage response of the battery to a current profile under different conditions. When the cell is in use, the model is used to track the internal state of the cell, providing an SoP estimate at any point in time through a prediction mechanism. Alternatively, a static model may be defined for relevant inputs from field measurements.
[0033] SoP estimation can be achieved by long DCIR tests involving current pulses with long relaxation times. This can be very time consuming and impractical for several reasons, one of which is that the resulting ECM must be calibrated after the battery has aged slightly. The use of EIS spectroscopy is more practical because it is relatively fast and does not involve stimulating the cell with large current pulses. It is much easier to recalibrate the ECM over the life of the battery because EIS measurements can be performed when the car is stationary (i.e. "turned off"). The concept of EIS-based ECM for SoP estimation is shown in Figure 2. Figure 1-4 shown.
[0034] Reference Figure 1 , a battery system 102 within a vehicle 100 is shown. The vehicle 100 may be an electric vehicle (EV) or a hybrid vehicle, such as a plug-in hybrid electric vehicle. The vehicle 100 may be considered a host of the battery system 102. The battery system 102 may include a BMS controller 110, a wireless battery management system (wBMS) 120, and a plurality of battery modules 130.
[0035] The BMS controller 110 may be a component of the vehicle 100 that is configured to interface with the wBMS 120. For example, the BMS controller 110 may be an electronic control unit (ECU) of the vehicle 100. The BMS controller 110 may execute a BMS controller application, which may be referred to as a safety application. For example, the BMS controller 110 may communicate with the wBMS 120 to receive information about the battery system, such as the state of charge, voltage, temperature, and any faults that have occurred. The BMS controller 110 may also communicate with other vehicle components, such as an inverter or a charger.
[0036] The wBMS 120 may be configured to interface between the battery module 130 and the BMS controller 110. For example, the wBMS 120 may receive packets including measurement messages and fault messages from the battery module 130. The wBMS 120 may aggregate messages from the individual battery modules to provide system-level information to the BMS controller 110. The wBMS 120 may include a wireless manager 122 and a wireless radio 124. The wireless manager 122 may be configured to generate messages for transmission to the battery module 130 and receive messages from the battery module 130. The wireless manager 122 may include a radio protocol stack. The wireless radio 124 may include one or more radios configured to send radio frequency (RF) signals to the battery module 130. The wireless radio 124 may be referred to as a head radio and may control (e.g., schedule) communications with the battery module 130.
[0037] The battery module 130 may include a wireless radio 140, a security processor 150, a battery monitoring system 160, and a plurality of batteries 170. The batteries 170 may be battery cells that store power. In some embodiments, each battery module 130 may include 3 to 24 individual battery cells 170.
[0038] The battery monitoring system 160 may be configured to monitor one or more parameters of a plurality of battery cells. For example, the battery monitoring system 160 may monitor the voltage and temperature of each battery. The battery monitoring system 160 may provide the measurements to the safety processor 150. The battery monitoring system 160 may be referred to as a battery monitoring integrated circuit (BMIC).
[0039] In various aspects, the battery monitoring system 160 includes an electrochemical impedance spectroscopy (EIS) system 125 for performing EIS measurements at various frequencies. The measurements may include, among other variables, voltage and current, from which an impedance spectrum for each battery may be calculated. This calculation may be performed by the safety processor 150 so that the measurements are used to calculate the SoP, thereby controlling how the battery is used. In this way, problems such as premature battery failure may be avoided.
[0040] The safety processor 150 may be a computer processor configured to execute computer code such as a script. In some implementations, the safety processor 150 is a separate processor connected to the wireless radio 140 and the battery monitoring system 160 via an interface such as a serial peripheral interface (SPI). In other implementations, the safety processor may be a processor of the wireless radio 140. The safety processor 150 may be configured to perform various tasks related to managing the battery module 130. For example, the safety processor 150 may be configured to execute a battery monitoring script to generate a battery information payload defined by a battery information payload format. The battery information payload may be based on a SoP calculated using EIS measurements from the EIS system 125. The safety processor may also receive commands (e.g., battery cell balancing commands) from the wBMS 120 and write the commands to the battery monitoring system 160. The safety processor 150 may include a script engine 152, a data processing engine 154, a scheduler 156, and a parser. The script engine 152 may execute a script 191 received from the wBMS 120. The data processing engine 154 can execute the data processing commands defined by the script to write the commands to the battery monitoring system 160 and process the data received from the battery monitoring system 160. In some implementations, the data processing engine can be a component of the script engine. The scheduler 156 can be configured to generate multiple packets according to the battery information payload format. For example, the scheduler 156 can operate according to the initialization schedule at initialization to potentially generate fault packets and operate according to the active mode schedule to generate measurement messages within the timing sub-loop. The parser 158 can be configured to parse various components of the script.
[0041] The wireless radio 140 may be configured to communicate with the wBMS 120 and / or the service device. The wireless radio 140 may be a radio similar to the wireless radio 124 configured to send and receive RF signals. The wireless radio 140 may be referred to as a tail radio because the wireless radio 140 may be managed by the wireless radio 124. The wireless radio 140 is configured to receive a container file 190 from the wBMS 120. The container file 190 includes metadata 192 that defines a battery information payload format and a battery monitoring script. The wireless radio 140 is configured to send at least the metadata 192 of the container file 190 to the service device. The wireless radio 140 is also configured to send a packet 181 including a battery information payload to the wBMS 120 or the service device.
[0042] To calculate the SoP, a battery model is used. In some aspects, the model is a battery equivalent circuit model (ECM). In other aspects, the model is a simplified model from a physics-based model. Other model representations of the battery may also be used while maintaining the spirit of the present disclosure.
[0043] Reference Figure 2 , a battery equivalent circuit model (ECM) 200 is shown according to an exemplary aspect. The battery ECM 200 includes an R portion 210 and an RC portion 220. Although one RC portion 220 is shown, Figure 5 As shown, other circuit models may have more than one RC section 220 .
[0044] The R section 210 of the ECM 200 includes a resistor R0 211. The RC section 220 includes a resistor R1 221 and a capacitor C1 222 connected in parallel and having a voltage VC1 therebetween. OCV The positive terminal of 230 is connected to one side of resistor R0211. Battery terminal voltage V term The positive terminal of is connected to one side of the RC part 220. The other side of the RC part 220 is connected to the other side of the resistor R0211. The battery terminal voltage V term The negative terminal is connected to the voltage V OCV 230's negative terminal.
[0045] Reference Figure 3 According to an exemplary aspect, it is shown Figure 2 A Nyquist plot 300 of the imaginary part of the electrochemical impedance spectrum of a battery versus the real part of the impedance of the battery ECM 200. Curve 310 represents the impedance calculated using the estimated ECM model of the battery and curve 320 represents the measured impedance of the battery.
[0046] Reference Figure 4 According to an exemplary aspect, it is shown Figure 2 A graph 400 of the terminal voltage of the battery ECM 200 versus time is shown. Specifically, time is shown on the x-axis, and the terminal voltage (V term ) is shown on the y-axis. Two curves are shown, a predicted curve 420 and a true curve 410.
[0047] Reference Figure 5 , shows another battery equivalent circuit model (ECM) 500 according to an exemplary aspect. The battery ECM 500 can be used to accommodate edge devices, such as Figure 1 In the wireless BMS shown and described. The battery ECM 500 includes an R section 510, a first RC section 520, and a second RC section 530. The R section 510 includes a resistor R0511. The first RC section 520 includes a resistor R1521 and a capacitor C1522 connected in parallel, and having a voltage V between them. C1 The second RC part 530 includes a resistor R2531 and a capacitor C2532 connected in parallel and having a voltage V between them. C2 Voltage V OCVThe positive terminal of 540 is connected to one side of resistor R0511. Battery terminal voltage V term The positive terminal of is connected to one side of the second RC part 530. The other side of the second RC part 530 is connected to the other side of the first RC part 520. The other side of the first RC part 520 is connected to the other side of the resistor R0511. The battery terminal voltage V term The negative terminal is connected to the voltage V OCV 540's negative terminal.
[0048] for Figure 5 For the ECM 500, the battery impedance is given by:
[0049]
[0050] Using EIS, this impedance is measured at various radial frequencies (ωk) and the ECM can be inferred from constrained nonlinear least squares estimates of the parameters.
[0051] refer to Figure 6-8 and Figure 10-11 , showing the estimated values of the ECM parameters as a function of temperature and state of charge (SoC) for the R-RC-RC approximation of a 3Ah Li-ion cell. These estimates were obtained using EIS scans performed at different temperatures and SoCs.
[0052] Fig. 9 The graph in shows the estimated cell impedance (Nyquist plot) compared to the measured EIS at different temperatures at 50% SoC. This clearly shows that the estimated ECM approximates the cell impedance very well. Specifically, Fig. 9 The Nyquist plot of the negative imaginary part of the battery impedance versus the real part of the battery impedance is shown, which in turn shows the impedance measured by EIS versus the impedance according to Figure 4 of the ECM estimated impedance, wherein according to an exemplary aspect, Figure 5 The ECM estimation parameters in were calculated as a function of temperature (T) using the nonlinear least squares method (NLLS);
[0053] EIS analysis uses a computer to find the model parameters that give the best agreement between the modeled impedance spectrum and the measured spectrum. A nonlinear least squares fit (NLLS) algorithm can be used. NLLS starts with an initial estimate of all the model parameters. From this initial point, the algorithm makes changes to a few or all parameter values and evaluates the resulting fit. If the change improves the fit, the new parameter value is accepted. If the change worsens the fit, the old parameter value is retained. Next, different parameter values are changed and the test is repeated. Each trial with a new value is called an iteration. Iterations continue until the goodness of fit exceeds the acceptance criteria, or until a limit is reached on the number of iterations.
[0054] Figure 6-11 The fit of the ECM using NLLS as a function of SoC and temperature is shown, and the circuit parameters are also shown as a function of the battery state (T, SoC, age). For example, the battery resistance decreases with increasing temperature, while the capacitance is roughly the opposite. To explain this change, an equivalent hypermodel of the battery can be inferred that maps the battery state to the ECM as follows:
[0055] R i =f R (T,SoC;θ i ).C i =f c (T,SoC;φ i ).
[0056] This super-model also helps us to perform efficient interpolation across different battery states without the need for explicit experiments.
[0057] Alternatively, projected gradient descent can be used to track ECM parameters by using EIS measurements. The gradient of the algorithm can be based on the estimated The partial derivatives of are derived as follows:
[0058]
[0059] The time domain voltage prediction technique will now be described according to exemplary aspects. To track the battery terminal voltage, a mathematical model is first derived based on the ECM parameters and the current applied to the battery as follows:
[0060]
[0061] Here, the component voltage dynamics are described by:
[0062]
[0063] Therefore, tracking V term This is equivalent to tracking OCV and V Ci . OCV is usually a nonlinear function of the SoC, such as Fig.12 As shown, the figure shows the OCV SoC of the battery, which evolves as follows:
[0064]
[0065] Where cap is the battery capacity. The ECM parameters also vary as nonlinear functions of the battery state. Taking this into account, an extended Kalman filter (EKF) is described to track V term .
[0066] A description will now be given regarding EKF-based terminal voltage tracking according to exemplary aspects.
[0067] The EKF technique involves forming a battery system matrix expressed as a function of a step index that depends on parameters of a hypermodel that in turn depends on the changing conditions.
[0068] Discretize the differential equation, Figure 5 The state space model of ECM 500 is as follows:
[0069]
[0070] in The measurement results are as follows:
[0071] y[n]=V term [n] = V OCV [n]([1 0 0]x[n])-[0 1 1]x[n]-R0i[n]
[0072] Therefore, when V term When measurements are available, the state estimate is corrected, and when V term When measurements are not available, the state vector can be predicted using a state space model. Consider the following state and measurement model:
[0073] x n =A n x n-1 +b n i n-1 +∈ n ,
[0074] y n =f(x n )+d n i n +δ n ,
[0075] The noise process is given by ∈ n ~N(0,Q),δ n ~N(0,R) gives a normal random vector. Then, Fig.13 As shown, the EKF with function 1300 can track the terminal voltage of the battery.
[0076] The EKF function 1300 includes a prediction portion 1310 , a linearization portion 1320 , and a correct portion 1330 .
[0077] It may be noted here that in addition to the EKF, other methods can be used to track and predict the battery state, including adaptive filtering and autoregressive models that predict the terminal voltage of the battery given the current state of the system. The method described here for calculating the power state using a super model of the battery is compatible with any such method.
[0078] A description will now be given of estimating the SoP of a battery according to exemplary aspects.
[0079] The SoP of a battery measures the maximum amount of charge that the battery can deliver. Two possible definitions and methods of calculating the SoP are described here. One definition is called constant current SoP and the other is called constant power SoP.
[0080] Regarding constant current SoP, a conservative estimate of the maximum power that can be extracted from the battery can be defined as follows:
[0081] P=V term,min i max
[0082] where i max is the maximum current pulse that can be extracted from the battery. The idea is that subject to the minimum terminal voltage V term,min Constraining the maximum battery current results in the maximum power that can be consumed by the battery.
[0083] Let V term (i,t+T p ) is the estimated value of the terminal voltage as a function of current i over time, where T p is the time period for calculating SoP. Then, the maximum value of i is the solution to the following problem:
[0084] V term (i,t+T p )-V term,nin =0
[0085] In one aspect, the Newton-Raphson method can be used to find the roots of this equation and thus calculate the SoP. However, this is a conservative estimate because it does not take into account the V term Instantaneous power fluctuations caused by changes.
[0086] Regarding constant power SoP, since it adapts to V term The maximum constant power that can be extracted from the battery is therefore more than the net energy output estimated by the constant current. Therefore, by solving for the time period of interest, T p The constant power P max The instantaneous current can more accurately estimate the maximum energy that can be extracted from the battery. This power is calculated by using the Powell conjugate direction method while taking i, Vterm Stay within the safe limits of the battery.
[0087] Unlike the aspects described herein, conventional techniques and corresponding methods for fitting ECMs are based on current pulse measurement techniques.
[0088] According to features of aspects described herein, EIS measurements can be used to train parameters of a generic algorithm model for SoP prediction in the field (when using an electric vehicle (EV)). In addition, EIS measurements are used to recalibrate parameters of the SoP algorithm model when the ignition is off (e.g., when the EV is parked in a garage). In addition, EIS measurements are used to dynamically recalibrate parameters of the SoP model in EV usage scenarios.
[0089] The main purpose of the aspects described herein is to estimate how much power a battery can output at any given time. The SoP can be indirectly quantified as the maximum allowed quiescent current that can be sustained for Δt seconds (or some pre-specified time period). "Maximum allowed" is defined as the maximum current that does not result in a constraint violation. Constraints can be defined for terminal voltage, current, state of charge, battery temperature, etc. Alternatively, the SoP can be indirectly quantified as the maximum allowed quiescent current that can be sustained for Δt seconds (or some pre-specified time period). "Maximum allowed" is defined as desired, for example, an amount that does not result in a temporary loss of capacity exceeding a specified fraction (e.g., 5%). The SoP (or maximum allowed current) is a dynamic variable that depends on the initial state of the battery at the time of prediction. In other words, any system that maps a measurable parameter to a SoP needs to include memory to achieve acceptable accuracy unless the battery state is provided as an additional input.
[0090] SoP estimation use cases include shutdown and startup. For shutdown, a model is developed to generate SoP predictions for an a priori given “battery state”. For example, the battery state can be defined as the most recent battery current, terminal voltage, open circuit voltage (OCV), state of charge (SoC), temperature, and impedance measurements (which may be different for individual cells and the entire battery pack). The SoP model is a multi-dimensional lookup table that maps each expected battery state to a corresponding SoP value (it can also be implemented as a function that performs this mapping / calculation when called, that is, not necessarily a hard-coded table read from memory when called).
[0091] For critical use cases, the model may be equipped with a tracker that starts from a known initial condition and then tracks the “state of the battery.” Whenever a SoP prediction is needed, the model can use the latest battery state provided by the tracker. The model’s SoP output is instantaneous (i.e., evaluated in real time) and only provides a prediction of the battery’s current state.
[0092] Reference Fig.14, illustrates a shutdown use case 1400 according to an exemplary aspect. The shutdown use case 1400 involves a state of power (SoP) model 1410, which inputs a measurable input (EIS), an expected battery state, and system constraints (e.g., maximum / minimum current / voltage (I / V) values), and outputs a predicted SoP.
[0093] In the key-off use case, the SoP map is generated offline, so no measurable data is required when generating the map (unless EIS is used to correct for aging factors, as shown above); however, even in the key-off use case, it is implicitly assumed that the SoP information will eventually be used for key-off purposes.
[0094] Reference Fig.15 , according to an exemplary aspect, a key use case 1500 is shown. The key use case 1500 involves a SoP model 1510 and a state tracker 1520. Both the SoP model 1510 and the state tracker 1520 input measurable inputs (EIS, V, I, time (T), SoC, state of health (SoH), etc.). The SoP model 1510 also inputs predicted battery states and system limits (e.g., maximum / minimum current / voltage (I / V) values) output from the state tracker 1520, and outputs a predicted SoP.
[0095] In the key-on use case, the battery status is tracked by a state tracker. In this case, more measurability is provided to the model along with system limits. Over time, the battery status that affects the SoP at a given time is tracked. One would recognize that the SoP is related to both the discharge and fast charge of the battery.
[0096] It should be recognized that the proposed model provides battery module-level information, while SoP is a battery pack-level metric; therefore, the individual module-level SoP measurements must be appropriately combined to produce a pack-level SoP.
[0097] The following are the SoP estimation methods for critical off and critical on use cases. For the critical use case, the regression model method involves training a baseline regression model at rest (offline) and generating SoP predictions by evaluating the regression function given predicted observables and battery state. The ECM method involves fitting the ECM to the EIS measurements at rest and generating SoP predictions by evaluating the ECM given predicted observables and battery state.
[0098] For key use cases, the regression model approach involves training a baseline regression model at rest (offline) and estimating the real-time SoP by evaluating the regression function given observables. The ECM with Kalman filter (KF) tracking approach involves fitting an ECM to EIS measurements at rest (offline), tracking the battery state using a KF, and evaluating predictions based on the ECM (with constraints) to infer the real-time SoP.
[0099] Return to reference Figure 2 And further reference Figure 16-17 According to an exemplary aspect, using an ECM to perform SoP estimation is shown.
[0100] As mentioned above Figure 2 The present invention relates to a battery ECM 200 used in one exemplary aspect. While maintaining the spirit of the present disclosure, other battery ECMs may also be used, such as Figure 5 The battery ECMs shown and described herein, as well as battery ECMs that would be readily contemplated by one of ordinary skill in the art given the teachings of the disclosure provided herein.
[0101] Fig.16 Shows Figure 2 FIG. 16 is a graph 1600 of the current versus time of the battery ECM 200. In particular, time is represented on the x-axis, and current (including I charge and I max ) is represented by the y-axis.
[0102] Fig.17 Shows Figure 2 1700 is a graph of the voltage of the battery ECM 200 over time. In particular, time is represented on the x-axis and voltage (including V charge and V max ) is represented by the y-axis. charge (Δt)=V max
[0103] First, if you use Figure 2 200, the battery is modeled using the R portion 210 and the RC portion 220 and the voltage source VOCV 230. Alternatively, another model or representation may be used, such as but not limited to Figure 5The ECM 500. The model is fit to the data taking into account parameter variations with temperature, SoC and age. The parameterization of the model can be in the form of a lookup table or a regression function. The voltage response to a current step of a desired duration (e.g. 10 s, 60 s, etc.) is calculated from an impedance spectrum determined at many frequencies, which can be used to fit a model characterizing the electrochemical performance of the cell. The parameterization can be done in closed form or by sequential reasoning with the KF algorithm. For improved accuracy, the ECM parameters can be adjusted (closed form approximation or online system identification in the KF framework). Constraints are set on current, SoC, voltage, etc. to find the maximum current; this gives P=I*V.
[0104] Reference Figure 18-19 , a method 1800 for offline pre-training of super model parameters according to an exemplary aspect is shown.
[0105] At step 1810, an EIS scan is performed on the battery pack at the selected temperature and SoC. This may result in an impedance spectrum including current and voltage measurements taken at different frequencies for each cell of the vehicle battery. The cells in the corpus have a set of operating characteristics similar to the vehicle battery. A battery pack having a set of operating characteristics similar to the vehicle battery is one having cells of similar cell chemistry, form factor, cell capacity, and operating conditions.
[0106] In one aspect, step 1810 may include one or more of steps 1810A-1810D.
[0107] Step 1810A includes, when the super model is an equivalent circuit model (ECM) super model, performing intelligent initialization of the parameter fitting method by setting the series resistor R in the R-RC ECM model to a minimum observed impedance value, connecting the series resistor R in series with one or more RC parallel sub-circuits, setting model parameters to determined values, and maintaining the determined values while sweeping within a certain range based on the value of the series resistor R to identify the determined values that minimize the objective function.
[0108] In step 1810B, the ECM is used as a hypermodel that maps various circuit elements (such as resistors, capacitors, inductors, and Warburg impedances of the ECM) to various states of the vehicle battery used in the EIS scan.
[0109] At step 1810C, a model simplified from the physics-based model is used as a hyper-model that maps the physical representation of the battery elements to the various states of the vehicle battery used in the EIS scan.
[0110] At step 1810D, an adaptive filter is used as a super model that uses the frequency response of the equivalent impedance of the vehicle battery learned using the EIS scans at various states of the vehicle battery to infer. In one aspect, the various states of the battery may include at least some of different temperature ranges of the battery, different states of charge (SoC) of the battery, the age of the battery, and the nature of the current load to which the battery is subjected.
[0111] In step 1820, the parameters of the super model are fit by applying optimization techniques to the results of the EIS scan. The super model includes a family of models, each of which defines the voltage response of a corresponding cell of a vehicle battery to a current curve under different conditions. The model family may include equivalent circuit models (ECMs), physical model models, or other types of battery models. The fitting can be performed using multidimensional lookup tables and / or regression functions. The results of the EIS scan can be applied to any one or more of the following optimization techniques: gradient-based linear optimization methods; non-negative least squares (NNLS); and convex optimization methods.
[0112] At 1830 , a state of power (SoP) of the vehicle battery is estimated using the hypermodel, results of the EIS scan, operating constraints of the vehicle battery, and various states of the vehicle battery used in the EIS scan.
[0113] Reference Fig. 20 , a method 2000 for calibrating and recalibrating a new battery is shown according to an exemplary aspect.
[0114] Prior to first use of the battery in a vehicle (e.g., at assembly), multiple electrochemical impedance spectroscopy (EIS) scans are performed on the battery at step 2010. This may result in an impedance spectrum including current and voltage measurements at different frequencies for each cell of the battery.
[0115] In step 220, the pre-trained super model of the battery is calibrated using the results of the EIS scan. The pre-trained super model includes a set of models, each of which defines the voltage response of a corresponding cell of the vehicle battery to a current curve under different conditions. The model family may include an equivalent circuit model (ECM), a physical model model, or other types of battery models.
[0116] At step 2030, the battery is cycled (normal use).
[0117] At step 2040, additional EIS scans are collected in situ in the key-off state (ie, after the battery is first used in a vehicle).
[0118] At step 2050, the pre-trained super model is recalibrated using the additional EIS scan results. In one aspect, for the recalibration, the respective complexity of the models in the family of models included in the super model may be increased as the battery age increases. In this way, the individual differences between each cell forming the battery may be accounted for, and their respective aging and corresponding effects may be taken into account.
[0119] It should be noted that according to certain aspects, EIS scans may also be collected during use (ie, key-on) rather than key-off.
[0120] Reference Figure 21-22 According to an exemplary aspect, it is further shown that Fig. 20 Step 2030 of method 2000.
[0121] In step 2131, the current output from each battery cell of the battery is measured.
[0122] At step 2132, a step is taken in the battery status tracker (if it is a power-on use case) to predict the current status of each cell of the battery.
[0123] At step 2133, the SoP of each cell of the battery is predicted in response to the current state of each cell of the battery.
[0124] In one aspect, step 2133 may include step 2133A.
[0125] At step 2133A, at least one constraint in the SoP prediction for each cell of the battery is considered. In one aspect, the at least one constraint may include at least one of a terminal voltage of the battery, a current of the battery, a temperature of the battery, and a state of charge (SoC) of the battery.
[0126] At step 2134, the SoP of each cell of the battery is combined to determine the battery SoP. It should be understood that various methods may be used to combine the individual cell SoPs into the battery SoP. Simple summation or more advanced combination techniques, including weighted summation, etc., may be used while maintaining the spirit of the present disclosure.
[0127] At step 2135, an action is performed on the battery in response to the single cell SoP and / or the battery SoP.
[0128] In one aspect, step 2135 may include one or more of steps 2035A and 2035B.
[0129] In step 2135A, the amount of current drawn from or put into the battery is controlled in response to the SoP having the lowest value among the battery SoP and / or the multi-cell SoPs of the plurality of batteries constituting the battery.
[0130] In step 2135B, service level actions are performed on the battery, including replacing the battery in response to the SoP being below a threshold and / or replacing a single battery having a single SoP below the same or another threshold to optimize the performance of a given battery by replacing its weak link (battery).
[0131] A further description of model fitting will now be given according to exemplary aspects.
[0132] In order to make the model applicable to impedance measurements at multiple frequencies Z n , the objective function is defined as follows:
[0133]
[0134] in is the complex-valued error of the fit, with the subscripts re and im indicating the real and imaginary parts. Its derivative with respect to the (real-valued) model parameters is given by:
[0135]
[0136] To simplify the derivation, the following shorthand notation is used:
[0137] T i =R i C i (9)
[0138] δ i =ωT i (10)
[0139]
[0140] The partial derivatives of the EIS model with respect to each parameter are as follows:
[0141]
[0142] The impedance model formula can also be abbreviated as follows:
[0143]
[0144] Given the formulas for these errors and their derivatives, nonlinear optimization techniques can be applied to tune the parameters to the dataset. It is imperative to ensure that the parameters remain non-negative.
[0145] A description will now be given regarding logarithmic scaling parameters according to exemplary aspects.
[0146] It is more natural to adjust the model parameters on a logarithmic scale, just as definition is just as simple. This has the added benefit of ensuring non-negativity. Conveniently, So applying the chain rule we get:
[0147]
[0148] A description will now be given of R-τ parameterization according to exemplary aspects.
[0149] This model can be expressed by τ i =R i C i Reparameterize. In this case, the impedance model is as follows:
[0150]
[0151] Its derivatives are as follows:
[0152]
[0153] A description will now be given regarding additional considerations according to exemplary aspects.
[0154] A line search at each iteration ensures that the error never increases. Additionally, well-known measures to improve convergence behavior such as adaptive step size and momentum can be implemented.
[0155] Initialization can be used. Initialization is particularly useful for R-RC type models to get close to the optimal value. R0 is set to the smallest observed impedance value. The τ parameters are then set to reasonable intermediate values (determined empirically / visually). Fixing these values, i The range of values of i>0 is scanned, all held equal, to find the value that minimizes the objective. Finally, R i and τ i (i>0) values are distributed deterministically in a small range around the initial value to avoid ambiguity. This seems to work well in practice. This leads to smart initialization of R-RC type models.
[0156] A further description of the supermodel will now be given according to exemplary aspects.
[0157] In practice, an impedance model that generalizes to many temperatures and SoCs is needed, since it is well known that the model parameters vary widely with respect to these states. If the RC models are subject to a standardized deterministic initialization procedure, the permutation problem between RC pairs can be avoided, which can severely complicate the generalization process. Once all individual models are fitted, a third-order polynomial is fit to each parameter via least squares regression.
[0158] A description will now be given regarding time-domain voltage prediction according to exemplary aspects.
[0159] Given an ECM, the time domain voltage response to a current input can be expressed. For an R-RC type model, the following voltage and current law equations can be used:
[0160]
[0161] The latter can be rewritten as follows:
[0162]
[0163] Therefore, we can predict V by simply solving for the voltage across the RC pair. term (t), and it is also desirable to discretize these voltage functions.
[0164] A description will now be given of the Euler approximation according to exemplary aspects.
[0165] Applying the positive Euler approximation, we obtain the following result:
[0166]
[0167] And applying Backward Euler gives the following result:
[0168]
[0169] Both solutions describe the voltage and voltage R j is a convex combination of i[n], but differs in the mixing coefficients and whether the current from the current or previous step is used. This form of difference equation, namely:
[0170] y[n]=ay[n-1]+(1-a)u[n] (34)
[0171] For a constant u[n], we have the following closed-form solution in the time domain:
[0172]
[0173] Under conditions of constant current and initial voltage, this equation can be used to predict the voltage at any time in the future.
[0174] A description will now be given of closed-form voltage prediction with respect to a discretized model according to exemplary aspects.
[0175] Substituting the RC voltage solution in the reverse Euler approximation into the terminal voltage formula, we get the following result:
[0176]
[0177] For a constant current i * , use and abbreviation:
[0178]
[0179] There are two problems with this expression:
[0180] SoC Dependency: It does not consider the dependency of ECM parameters on SoC, which may change non-negligibly during the forecast period depending on the current magnitude. This is addressed by SoC-dependent ECM, due to c j is a constant, so the equation does not apply to different SoCs.
[0181] Initial conditions: The voltage on each RC pair must be Provides an initial value. This is handled by an appropriate tracking scheme.
[0182] A description will now be given of explicit solutions to continuous-time models according to exemplary aspects.
[0183] Since the entire system model is fully decoupled and each ODE is linear first order, the system can be explicitly solved to obtain the voltage That is, the following result is directly obtained from (27):
[0184]
[0185] Assuming that i(t) = i0 is constant in the interval of interest, we obtain the following result:
[0186]
[0187] Similarly, if the system parameter R j and τ j are time-varying (for example, if they depend on SOC or battery temperature), then The explicit solution of can be written as follows:
[0188]
[0189] where φ=φ(t) is The antiderivative of If the model parameter R j (t) and can be explicitly described as a function of time, then the above integral can be solved either explicitly or numerically. For example, given the initial SOC and battery temperature, it should be possible to approximate the model parameters with a low-order polynomial (e.g., linear or quadratic) within the small time window considered.
[0190] For example, R j (t) and τ j (t) can be modeled as:
[0191] (1) For t∈[t0, t0+ε] and all j, R j (t) = a j t+b j , τ j =a j t+β j
[0192] (2) For t∈[t0, t0+∈] and all j, R j (t) = a j t 2 +b j , τ j =a j t+β j
[0193] where ∈ is a small positive constant.
[0194] A description will now be given regarding prediction based on explicit solutions of continuous-time state-space models according to exemplary aspects.
[0195] When the battery is in use, the SoP algorithm must be able to provide an estimate of the maximum power that can be put into or extracted from the battery within some predefined small time window at any moment in time. Therefore, in order to accurately estimate P(t), it is necessary to calculate the maximum power that can be put into or extracted from the battery within the time range [t, t+T p ]Battery terminal voltage waveform V term (·) for accurate prediction. Battery performance can be evaluated by the discharge or charge energy E(t) while maintaining SoP. The energy is expressed as follows:
[0196]
[0197]
[0198] The energy can be approximated by the composite Simpson's rule Riemann sum shown below:
[0199]
[0200] Where j = 0, 1, ..., N-1, N is N even subintervals, and the j-th time step is:
[0201] For [t, t+T p ],
[0202] A description will now be given regarding conservative SoP estimation according to exemplary aspects.
[0203] Due to the change in battery voltage, maxApplying a constant discharge or charge at V will not result in a constant power. Using the average power as the SoP for a constant power discharge or charge may not be possible because the term,min The voltage will need to be greater than i max For higher currents, this may be limited by the manufacturer. The peak power is given in (45), which is an underestimate of the maximum power.
[0204] P=V term,min i max (45)
[0205] A constant current pulse means that the rate of change of the current is infinite at the beginning of the pulse. This can be considered a worst-case scenario and is used for the following reasons:
[0206] -The current can be easily controlled by power electronics.
[0207] -The voltage variation during the constant current pulse is negligible.
[0208] A description will now be given of a Newton-Raphson method for finding i for a conservative SoP estimate according to exemplary aspects. max .
[0209] The maximum battery current results in the maximum power that can be extracted from the battery and is subject to the minimum terminal voltage V term,min The problem statement becomes:
[0210] Problem: Solve for the given V term (i,t+T p ) max =i * , so that f(i * )=V term (i * ,t+T p )-V term,min =0 in the time range t[,t+T p ]Inside.
[0211] The Newton-Raphson method is a powerful technique for numerically solving root-finding problems using linear approximations. max are two initial guesses for i0 and i1, use the secant method to approximate the derivative, and follow (46) until [i n -i n-1 ] satisfies the termination tolerance.
[0212]
[0213] V term (i,t+T p )The function returns t+T pThe final terminal voltage at time T is predicted using the state space model in (56) without using the observed value of the constant current i. p status.
[0214] The main issues with this approach include:
[0215] - Wrong initial guess that the function does not converge;
[0216] - Sub-quadratic convergence rate; and
[0217] - at the previous time step t+T p The noise voltage may be lower than V term,min solution.
[0218] Possible improvements:
[0219] -Use Brent's method combining bisection, secant and inverse quadratic interpolation to confine the current to an interval;
[0220] - Log updated to limit current to non-negative values; and
[0221] - for f(i n ) use a termination condition instead of i for steep slopes n
[0222] A method for finding the SoP will now be presented according to exemplary aspects. max Description of the constant power method.
[0223] The maximum constant power produces more energy output than the conservative estimate of SoP. The Powell method is used to numerically calculate that at t+T p The maximum constant power discharge that results in the specified minimum terminal voltage of the battery at t. The battery has maximum current discharge and charge pulses that need to be considered to limit the current profile. This is done by solving for t+T p If the current reaches a value higher than the cut-off voltage, the current is equal to T p However, as voltage, temperature, and SoC change, it is necessary to find a current distribution that results in constant power discharge.
[0224] A description will now be given regarding constant power emulation according to exemplary aspects.
[0225] To find the maximum constant power discharge or charge, the current profile needs to be solved to compensate for the change in model voltage. The current at each time step is calculated numerically using Powell's method. In shorthand form
[0226]
[0227] At the next time step n, 26 and 33 can be used to derive the current distribution.
[0228] p[n]=V term [t]i[n] (48)
[0229]
[0230]
[0231]
[0232] V OCV Depends on SoC and temperature. To solve for the current i[n] that results in a constant power discharge, the inverse OCV-SoC relationship can be used to convert V OCV [n] is represented by i[n] and V OCV [n-1]. However, since the OCV-SoC function has not yet been defined, V can be approximated with a sufficiently small time step due to negligible SoC changes. OCV [n]≈V OCV [n-1]. The current i[n] can be solved directly using (51), which is a quadratic equation with a negative sign for a positive total battery voltage.
[0233]
[0234] The results show that the required power is very close to the computational power for sufficiently small time steps (~100ms). This capability can be used to more efficiently compute SoP on embedded systems.
[0235] A description of tracking will now be given according to exemplary aspects. The description will involve state space models and EKF approximation.
[0236] A description of the state-space model will now be given according to exemplary aspects.
[0237] The full state-space model corresponding to the inverse Euler discretization of 2 RC pairs has the state transfer equations:
[0238]
[0239] It can be abbreviated as follows:
[0240] x[n]=Ax[n-1]+bi[n] (55)
[0241] For a Jx1 state vector x, a J×J state transfer matrix A, and a Jx1 control input vector b. The following abbreviations are used:
[0242] The measurement formula is as follows:
[0243] V term [n] = V OCV [n]([1 0 0]x[n])-[0 1 1]x[n]R0i[n] (56)
[0244] The abbreviations are as follows:
[0245] y[n]=f(x[n])+di[n] (57)
[0246] For the nonlinear function f(·) and the control input scalar d, the dependence of the SoC (i.e., s[n]) on the OCV at time step n has been included. The OCV-SoC function f(·) can be estimated empirically and approximated with a 9th-order polynomial, for example, to obtain the EKF approximation.
[0247] The presence of the control input i[n] implies very simple adjustments to the prediction and correction steps of the Kalman filter.
[0248] A description of the EKF approximation will now be given according to exemplary aspects.
[0249] The Extended Kalman Filter (EKF) is a useful and simple method for approximating nonlinearities in dynamic models. Given the state-space model defined above, assume the following generative model:
[0250] x n =A n x n-1 +b n i n +∈ n (58)
[0251] Y n =fx n +d n i n +δ n (59)
[0252] Noise processing:
[0253] ∈~N(0,Q) (60)
[0254] δ~N(0,R) (61)
[0255] The corresponding EKF equation is as follows: Fig.13 As shown:
[0256] predict
[0257]
[0258] Linearization
[0259]
[0260] Correction
[0261]
[0262]
[0263] The state-dependent measurement function is used
[0264] The system matrices are expressed as functions of the step indices, since they depend on the ECM parameters, which themselves are functions of the SoC and temperature. The SoC is tracked in the model and the temperature is provided externally, so the nonlinearity introduced by this interdependence is assumed to be negligible. The system matrices also depend on the step duration, which varies with time.
[0265] The EKF will only perform the prediction step when no measurements are available. This allows it to continue "tracking" when measurements are unavailable.
[0266] Additional aspects
[0267] The present disclosure may additionally include one or more of the following aspects.
[0268] Aspect 1. A method for pre-training a state of power (SoP) configured to predict a vehicle battery, the method comprising:
[0269] performing electrochemical impedance spectroscopy (EIS) scans on a plurality of batteries having a set of operating characteristics similar to the vehicle battery, the EIS scans being performed at various states of the vehicle battery; and
[0270] Parameters of the super model are fitted by applying an optimization technique to the results of the EIS scan, the super model comprising a family of models, each model defining a voltage response of a corresponding battery cell among a plurality of battery cells of the vehicle battery to a current distribution under various states of the vehicle battery.
[0271] Aspect 2. The method according to aspect 1 further includes estimating the state of power (SoP) of the vehicle battery using the hypermodel, results of the EIS scan, operating constraints of the vehicle battery, and various states of the vehicle battery used in the EIS scan.
[0272] Aspect 3. A method according to Aspect 1, wherein the various states of the vehicle battery include at least some of the following: different temperature ranges of the vehicle battery; different states of charge (SoC) of the vehicle battery; the life of the vehicle battery; and the nature of the current load borne by the vehicle battery.
[0273] Aspect 4. A method according to aspect 1, wherein the super model is an equivalent circuit model (ECM), and the super model is an equivalent circuit model (ECM) that maps various circuit elements of the ECM, such as resistors, capacitors, inductors, and Warburg impedances, to various states of the vehicle battery used in the EIS scan.
[0274] Aspect 5. The method of aspect 1, wherein the hypermodel is a model simplified from a physics-based model that maps physical representations of battery elements to various states of a vehicle battery used in an EIS scan.
[0275] Aspect 6. The method according to aspect 1, wherein the super model is an adaptive filter inferred using a frequency response of an equivalent impedance of a vehicle battery learned using EIS scans at various states of the vehicle battery.
[0276] Aspect 7. The method according to any one of aspects 1 to 6, wherein the fitting is performed offline.
[0277] Aspect 8. The method according to any one of aspects 1 to 7, wherein the hypermodel, expected battery state, results of the EIS scan, and battery system current and voltage limits are mapped into corresponding SoP values using a multi-dimensional lookup table.
[0278] Aspect 9. The method according to aspect 1, wherein mapping from the EIS scan, the various states of the vehicle battery, and the operating constraints of the vehicle battery to the power state (SoP) of the battery is performed using a regression function learned at various states of the vehicle battery.
[0279] Aspect 10. The method according to aspect 1, wherein the results of the EIS scan are used to infer a super model using an optimization technique selected from the group consisting of a gradient-based linear optimization method, a non-negative least squares (NNLS) method, and a convex optimization method.
[0280] Aspect 11. A method according to aspect 1, wherein the super model is an equivalent circuit model (ECM) family, and the method also includes performing an intelligent initialization of a parameter fitting method by setting a series resistor R in an R-RC ECM model to a minimum observed impedance value, wherein the R-RC ECM model has a series resistor R in series with one or more RC parallel sub-circuits; setting a model parameter to a determined value and maintaining the determined value unchanged when scanning within a certain range based on the value of the series resistor R to identify the determined value that minimizes the objective function.
[0281] Aspect 12. A method for predicting a state of power (SoP) of a battery, the method comprising:
[0282] performing multiple electrochemical impedance spectroscopy (EIS) scans on the battery prior to first use of the battery in a vehicle;
[0283] calibrating a pre-trained supermodel using results from the plurality of EIS scans, the supermodel comprising a family of models, each model defining a voltage response of a corresponding cell among a plurality of cells of a battery to a current distribution at various states of the battery;
[0284] performing a plurality of additional EIS scans on the battery after the battery is first used in the vehicle;
[0285] recalibrating the pre-trained supermodel using results from a plurality of additional EIS scans;
[0286] predicting a SoP for each of a plurality of batteries of the vehicle battery in response to a current battery state of each of the plurality of batteries;
[0287] combining the SoPs of each of the plurality of battery cells into a battery SoP; and
[0288] An amount of current drawn from or put into the battery is controlled in response to the SoP value.
[0289] Aspect 13. The method of aspect 12, wherein the plurality of additional EIS scans are performed periodically after initial use of the battery in the vehicle.
[0290] Aspect 14. The method according to aspect 12, wherein the various states of the battery include at least some of the following: different temperature ranges of the battery, different states of charge (SoC) of the battery, the age of the battery, and the nature of the current load borne by the battery.
[0291] Aspect 15. The method according to aspect 12, wherein the super model is an equivalent circuit model (ECM), mapping various circuit elements of the ECM, such as resistors, capacitors, inductors, and Warburg impedances, to various states of the battery used in the EIS scan.
[0292] Aspect 16. The method of aspect 12, wherein the hypermodel is a model simplified from a physics-based model that maps physical representations of battery elements to various states of the battery used in an EIS scan.
[0293] Aspect 17. The method according to aspect 12, wherein the super model is an adaptive filter inferred using the frequency response of the equivalent impedance of the battery learned using EIS scans at various states of the battery.
[0294] Aspect 18. The method according to aspect 12, wherein the results of the EIS scan are used to infer a super model using an optimization technique selected from the group consisting of a gradient-based linear optimization method, a non-negative least squares (NNLS) method, and a convex optimization method.
[0295] Aspect 19. A method according to Aspect 12, wherein the super model is an equivalent circuit model (ECM) family, and the method also includes performing an intelligent initialization of a parameter fitting method by setting a series resistor R in an R-RC ECM model to a minimum observed impedance value, wherein the R-RC ECM model has a series resistor R in series with one or more RC parallel sub-circuits; setting a model parameter to a determined value and keeping the determined value unchanged when scanning within a certain range based on the value of the series resistor R to identify the determined value that minimizes the objective function.
[0296] Aspect 20. The method of aspect 12, wherein the plurality of additional EIS scans are performed while the battery is in a key-on state.
[0297] Aspect 21. The method according to aspect 12 or 20, further comprising: while the battery is in a key-on state:
[0298] measuring a current output of each of a plurality of battery cells of the battery; and
[0299] A present battery state of each of the plurality of battery cells of the battery is determined in response to the current output from each of the plurality of battery cells of the battery.
[0300] Aspect 22. The method according to aspect 12 further includes predicting the SoP of at least one of the battery cells of the battery, wherein at least one constraint includes at least one of a terminal voltage of the battery, a current of the battery, a temperature of the battery, and a state of charge (SoC) of the battery.
[0301] Aspect 23. The method of aspect 12, wherein the plurality of additional EIS scans are performed to update the super model while the battery is in a key-off state.
[0302] Aspect 24. The method of aspect 12, wherein in a key-off state, the battery SoP is estimated using a regression algorithm given the EIS measurement value and the current state of the battery.
[0303] Aspect 25. The method of aspect 12, wherein the SoP includes a maximum allowed quiescent current that can be sustained over a given period of time.
[0304] Aspect 26. The method of aspect 25, wherein the maximum allowed quiescent current comprises a current level that does not result in a constraint violation.
[0305] Aspect 27. The method of aspect 25, wherein the maximum allowed quiescent current comprises a current level that does not cause a temporary loss of capacity exceeding a specified fraction within a given time period.
[0306] Aspect 28. The method of aspect 12, wherein the SoP comprises a maximum allowed constant power that can be sustained over a given period of time.
[0307] Aspect 29. The method of aspect 20, further comprising tracking the battery state starting from a known initial condition in response to a measurable input comprising an EIS scan result.
[0308] Aspect 30. The method according to Aspect 20 further includes tracking a battery state including a terminal voltage of the battery using an extended Kalman filter (EKF) technique, wherein the EKF technique includes predicting a battery state vector using a state space model and predicting a terminal voltage of a battery response within a desired time period.
[0309] Aspect 31. A method according to aspect 30, wherein the EKF technique includes forming a battery system matrix, the battery system matrix being represented as a function of a step index, the step index depending on parameters of the hypermodel, which in turn depend on various states of the battery.
[0310] Aspect 32. The method according to aspect 20 further comprises estimating a battery state including a battery terminal voltage using a closed-form solution of a discrete approximation of a state-space model and predicting the terminal voltage of the battery response over a desired time period.
[0311] Aspect 33. The method of aspect 20, further comprising generating a prediction of a current SoP of the battery in response to the current prediction of the battery state, measurable inputs, and system constraints related to current and voltage maximum and minimum values.
[0312] Aspect 34. The method according to aspect 33, wherein the battery status includes a current battery current, a current terminal voltage, a current open circuit voltage (OCV), and a current impedance measurement value obtained using EIS.
[0313] Aspect 35. The method according to aspect 12, wherein, for recalibration, the respective complexity of the models in the family of models included in the super model increases as the battery life increases.
[0314] Aspect 36. The method of aspect 12, wherein the SoP of the battery is equal to one of the SoPs of each of the plurality of battery cells.
[0315] Aspect 37. The method according to aspect 12, wherein the SoP value of the battery is equal to a lowest value among a plurality of battery cell SoPs of a plurality of battery cells constituting the battery.
[0316] Aspect 38. The method of aspect 12, wherein the SoP value is equal to the battery SoP.
[0317] Aspect 39: A system for predicting a state of power (SoP) of a battery, the system comprising:
[0318] an electrochemical impedance spectroscopy (EIS) system for performing a plurality of EIS scans on a battery prior to first use of the battery in a vehicle and performing a plurality of additional EIS scans on the battery in the vehicle;
[0319] a storage device for storing program code; and
[0320] A processing device, operatively coupled to the EIS system and the memory device, for executing program code to:
[0321] calibrating a pre-trained super model using results from the plurality of EIS scans, the pre-trained super model comprising a family of models, each model defining a voltage response of a corresponding cell among a plurality of cells of the battery to a current profile at various states of the battery;
[0322] recalibrating the pre-trained supermodel using results from a plurality of additional EIS scans;
[0323] predicting a SoP of each of a plurality of battery cells of the vehicle battery in response to a current battery state of each of the plurality of battery cells;
[0324] combining the SoPs of each of the plurality of battery cells into a battery SoP; and
[0325] An amount of current drawn from or put into the battery is controlled in response to at least one of the SoP of each of the plurality of battery cells or the battery SoP.
[0326] Aspect 40. A system having one or more components configured to perform the functions described in any one of aspects 13 to 38.
[0327] It should be noted that all specifications, dimensions and relationships (e.g., number of components, operations, steps, etc.) described herein are for example and teaching purposes only. Such information may be significantly changed without departing from the spirit of the present disclosure or the scope of the appended claims. These specifications apply only to one non-limiting example, and therefore, they should be interpreted as such. In the foregoing description, exemplary aspects have been described with reference to specific component arrangements. Various modifications and changes may be made to these aspects without departing from the scope of the appended claims. Therefore, the specification and drawings should be regarded as illustrative, not restrictive.
[0328] Note that through many of the examples provided herein, interactions may be described in terms of two, three, four or more electrical components. However, this is done for clarity and example only. It should be understood that the system may be integrated in any suitable manner. Along similar design alternatives, any of the components, modules, and elements shown in the drawings may be combined in a variety of possible configurations, all of which are clearly within the broad scope of this specification. In some cases, it may be easier to describe one or more functions of a given set of flows by referencing only a limited number of electrical components. It should be understood that the drawings and the circuits they teach are easily expandable and can accommodate a large number of components and more complex / sophisticated arrangements and configurations. Therefore, the examples provided should not limit the scope of the circuit or inhibit its broad teachings that may be applied to countless other architectures.
[0329] It should also be noted that in this specification, references to various features (e.g., elements, structures, modules, components, steps, operations, features, etc.) included in "one aspect", "exemplary aspect", "on one hand", "on the other hand", "some aspects", "various aspects", "other aspects", "alternative aspects", etc.) are intended to mean that any such features are included in one or more aspects of the present disclosure, but may or may not be combined in the same aspect.
[0330] It should be understood that, for example, in the case of "A / B", "A and / or B", and "at least one of A and B", the use of any of the following " / ", "and / or", is intended to include selecting only the first listed option (A), or selecting only the second listed option (B), or selecting both options (A and B). As another example, in the case of "A, B, and / or C" and "at least one of A, B, and C", such language is intended to include selecting only the first listed option (A), or selecting only the second listed option (B), or selecting only the third listed option (C), or selecting only the first and second listed options (A and B), or selecting only the first and third listed options (A and C), or selecting only the second and third listed options (B and C), or selecting all three options (A, B, and C). As will be readily appreciated by those of ordinary skill in this and related arts, this can be extended to any number of items listed.
[0331] It should also be noted that the functions related to the circuit architecture only illustrate some possible circuit architecture functions that can be performed by or within the system shown in the figure. Where appropriate, some of these operations may be deleted or removed, or these operations may be significantly modified or changed without departing from the scope of the present disclosure. In addition, the timing of these operations may vary greatly. The above-mentioned operation flow is for example and discussion purposes only. The aspects described herein provide great flexibility because any suitable arrangement, timing, configuration, and timing mechanism can be provided without departing from the teachings of the present disclosure.
[0332] Numerous other variations, substitutions, changes, and modifications may be ascertained by those skilled in the art, and the present disclosure is intended to embrace all such variations, substitutions, changes, and modifications that fall within the scope of the appended claims.
[0333] Note that all optional features of the above-described devices and systems may also be implemented with respect to the methods or processes described herein, and that details in the examples may be used anywhere in one or more aspects.
[0334] In the above cases, the means for "for" may include (but is not limited to) using any suitable components discussed herein, as well as any suitable software, circuits, hubs, computer code, logic, algorithms, hardware, controllers, interfaces, links, buses, communication paths, etc.
[0335] Note that for the above examples, as well as many other examples provided herein, interactions may be described in terms of two, three, or four network elements. However, this is done for clarity and example purposes only. In some cases, it may be easier to describe one or more functions of a given set of flows by referencing only a limited number of network elements. It should be understood that the topologies shown and described with reference to the accompanying drawings (and their teachings) are readily scalable to accommodate a large number of components and more complex / complex arrangements and configurations. Therefore, the examples provided should not limit the scope of the topologies shown or inhibit their broad teachings that may be applied to countless other architectures.
[0336] It is also important to note that the steps in the preceding flowcharts illustrate only some of the possible signaling scenarios and patterns that may be performed by or within the communication system shown in the figures. Where appropriate, some of these steps may be deleted or removed, or may be substantially modified or changed without departing from the scope of the present disclosure. In addition, many of these operations are described as being performed simultaneously or in parallel with one or more additional operations. However, the timing of these operations may vary significantly. The above operational flows are provided for example and discussion purposes only. The communication system shown in the accompanying drawings provides great flexibility in that any suitable arrangement, timing, configuration, and timing mechanism may be provided without departing from the teachings of the present disclosure.
[0337] Although the present disclosure has been described in detail with reference to specific arrangements and configurations, these example configurations and arrangements may be significantly changed without departing from the scope of the present disclosure. For example, although the present disclosure has been described with reference to a specific communication exchange, aspects described herein may be applicable to other architectures.
[0338] Numerous other variations, substitutions, changes, and modifications may be ascertained by those skilled in the art, and the present disclosure is intended to encompass all such variations, substitutions, changes, and modifications that fall within the scope of the appended claims. To assist the United States Patent and Trademark Office (USPTO) and any reader of any patent issued in this application in interpreting the claims appended to this application, applicants wish to note that applicants: (a) do not intend to invoke 35 U.S.C. § 142, paragraph 6(6) as it exists on the date of filing this application, unless “means” or “step” is specifically used in a particular claim; and (b) do not intend, by any statement in the specification, to limit the present disclosure in any manner not reflected in the appended claims.
Claims
1. A method for pre-training a state of power (SoP) configured to predict a vehicle battery, the method comprising: performing electrochemical impedance spectroscopy (EIS) scans on a plurality of batteries having a set of operating characteristics similar to the vehicle battery, the EIS scans being performed at various states of the vehicle battery; and Parameters of the super model are fitted by applying an optimization technique to the results of the EIS scan, the super model comprising a family of models, each model defining a voltage response of a corresponding battery cell among a plurality of battery cells of the vehicle battery to a current distribution under various states of the vehicle battery.
2. The method of claim 1 , further comprising estimating a state of power (SoP) of the vehicle battery using the hypermodel, results of the EIS scan, operating constraints of the vehicle battery, and various states of the vehicle battery used in the EIS scan.
3. The method of claim 1 , wherein the various states of the vehicle battery include at least some of: different temperature ranges of the vehicle battery; different states of charge (SoC) of the vehicle battery; the age of the vehicle battery; and the nature of the current load to which the vehicle battery is subjected.
4. The method of claim 1 , wherein the supermodel is an equivalent circuit model (ECM) that maps various circuit elements of the ECM, such as resistors, capacitors, inductors, and Warburg impedances, to various states of a vehicle battery used in an EIS scan.
5. The method of claim 1, wherein the hypermodel is a model simplified from a physics-based model that maps physical representations of battery components to various states of a vehicle battery used in an EIS scan. 6 . The method of claim 1 , wherein the super model is an adaptive filter inferred using a frequency response of an equivalent impedance of a vehicle battery learned using EIS scans at various states of the vehicle battery.
7. The method according to any one of claims 1 to 6, wherein the fitting is performed off-line.
8. The method of any one of claims 1-7, wherein the hypermodel, expected battery state, results of an EIS scan, and battery system current and voltage limits are mapped into corresponding SoP values using a multi-dimensional lookup table.
9. The method of claim 1, wherein mapping from the EIS scan, the various states of the vehicle battery, and the operating constraints of the vehicle battery to the state of power (SoP) of the battery is performed using a regression function learned at various states of the vehicle battery.
10. The method of claim 1, wherein the results of the EIS scan are used to infer a hypermodel using an optimization technique selected from the group consisting of a gradient-based linear optimization method, a non-negative least squares (NNLS) method, and a convex optimization method.
11. The method of claim 1 , wherein the supermodel is an equivalent circuit model (ECM) family, and the method further comprises: An intelligent initialization of a parameter fitting method is performed by setting a series resistor R in an R-RC ECM model having a series resistor R in series with one or more RC parallel subcircuits to a minimum observed impedance value; setting model parameters to determined values and maintaining the determined values while sweeping within a range based on the value of the series resistor R to identify the determined values that minimize an objective function.
12. A method for predicting a state of power (SoP) of a battery, the method comprising: performing multiple electrochemical impedance spectroscopy (EIS) scans on the battery prior to first use of the battery in a vehicle; calibrating a pre-trained supermodel using results from the plurality of EIS scans, the supermodel comprising a family of models, each model defining a voltage response of a corresponding cell among a plurality of cells of a battery to a current distribution at various states of the battery; performing a plurality of additional EIS scans on the battery after the battery is first used in the vehicle; recalibrating the pre-trained supermodel using results from a plurality of additional EIS scans; predicting a SoP for each of a plurality of batteries of the vehicle battery in response to a current battery state of each of the plurality of batteries; combining the SoPs of each of the plurality of battery cells into a battery SoP; as well as An amount of current drawn from or put into the battery is controlled in response to the SoP value.
13. The method of claim 12, wherein the plurality of additional EIS scans are performed periodically after initial use of the battery in the vehicle.
14. The method of claim 12, wherein the various states of the battery include at least some of the following: different temperature ranges of the battery, different states of charge (SoC) of the battery, age of the battery, and nature of the current load experienced by the battery.
15. The method of claim 12, wherein the supermodel is an equivalent circuit model (ECM) mapping various circuit elements of the ECM, such as resistors, capacitors, inductors, and Warburg impedances, to various states of the battery used in the EIS scan.
16. The method of claim 12, wherein the hypermodel is a model simplified from a physics-based model that maps physical representations of battery components to various states of the battery used in an EIS scan.
17. The method of claim 12, wherein the super model is an adaptive filter inferred using a frequency response of an equivalent impedance of a battery learned using EIS scans at various states of the battery.
18. The method of claim 12, wherein the results of the EIS scan are used to infer a hypermodel using an optimization technique selected from the group consisting of a gradient-based linear optimization method, a non-negative least squares (NNLS) method, and a convex optimization method.
19. The method of claim 12, wherein the supermodel is an equivalent circuit model (ECM) family, and the method further comprises performing an intelligent initialization of a parameter fitting method by setting a series resistor R in an R-RC ECM model having a series resistor R in series with one or more RC parallel subcircuits to a minimum observed impedance value; setting model parameters to determined values and maintaining the determined values while sweeping within a certain range based on the value of the series resistor R to identify the determined values that minimize an objective function.
20. The method of claim 12, wherein the plurality of additional EIS scans are performed while the battery is in a key-on state.
21. The method of claim 12 or 20, further comprising: during a key-on state of the battery: measuring a current output of each of a plurality of battery cells of the battery; and A present battery state of each of the plurality of battery cells of the battery is determined in response to the current output from each of the plurality of battery cells of the battery.
22. The method of claim 12, further comprising predicting the SoP of at least one of the battery cells of the battery, wherein the at least one constraint comprises at least one of a terminal voltage of the battery, a current of the battery, a temperature of the battery, and a state of charge (SoC) of the battery.
23. The method of claim 12, wherein the plurality of additional EIS scans are performed to update the hyper-model while the battery is in a key-off state.
24. The method of claim 12, wherein in a key-off state, the battery SoP is estimated using a regression algorithm given the EIS measurements and the current state of the battery.
25. The method of claim 12, wherein the SoP includes a maximum allowed quiescent current that can be sustained over a given period of time.
26. The method of claim 25, wherein the maximum allowed quiescent current comprises a current level that does not result in a constraint violation.
27. The method of claim 25, wherein the maximum allowed quiescent current comprises a current level that does not cause a temporary loss of capacity exceeding a specified fraction over a given period of time.
28. The method of claim 12, wherein the SoP comprises a maximum allowed constant power that can be sustained over a given period of time.
29. The method of claim 20, further comprising tracking the battery state starting from a known initial condition in response to a measurable input comprising an EIS scan result.
30. The method of claim 20 further comprises tracking a battery state including a terminal voltage of the battery using an extended Kalman filter (EKF) technique, the EKF technique comprising predicting a battery state vector using a state space model and predicting a terminal voltage of a battery response over a desired time period.
31. The method of claim 30, wherein the EKF technique includes forming a battery system matrix represented as a function of step indices that depend on parameters of the hypermodel that in turn depend on various states of the battery.
32. The method of claim 12, further comprising, under a key-on condition, estimating a battery state including a battery terminal voltage using a closed-form solution of a discrete approximation of a state-space model and predicting the terminal voltage of the battery response over a desired time period.
33. The method of claim 20, further comprising generating a prediction of a current SoP for the battery in response to the current prediction of the battery state, measurable inputs, and system constraints related to current and voltage maximum and minimum values.
34. The method of claim 33, wherein the battery status comprises present battery current, present terminal voltage, present open circuit voltage (OCV), and present impedance measurements obtained using EIS.
35. The method of claim 12, wherein for recalibration, the respective complexity of the models in the family of models included in the super-model increases as battery life increases.
36. The method of claim 12, wherein the SoP of the battery is equal to one of the SoPs of each of a plurality of battery cells.
37. The method of claim 12, wherein the SoP value of the battery is equal to a lowest value among a plurality of battery cell SoPs of a plurality of battery cells constituting the battery.
38. The method of claim 12, wherein the SoP value is equal to the battery SoP.
39. A system for predicting the state of power (SoP) of a battery, the system comprising: an electrochemical impedance spectroscopy (EIS) system for performing a plurality of EIS scans on a battery prior to first use of the battery in a vehicle and performing a plurality of additional EIS scans on the battery in the vehicle; a storage device for storing program code; and A processing device, operatively coupled to the EIS system and the memory device, for executing program code to: calibrating a pre-trained super model using results from the plurality of EIS scans, the pre-trained super model comprising a family of models, each model defining a voltage response of a corresponding cell among a plurality of cells of the battery to a current profile at various states of the battery; recalibrating the pre-trained supermodel using results from a plurality of additional EIS scans; predicting a SoP of each of a plurality of battery cells of the vehicle battery in response to a current battery state of each of the plurality of battery cells; combining the SoPs of each of the plurality of battery cells into a battery SoP; as well as An amount of current drawn from or put into the battery is controlled in response to at least one of the SoP of each of the plurality of battery cells or the battery SoP.
40. The system of claim 39, wherein the system is included in a battery management system.
41. The system of claim 39, wherein the plurality of additional EIS scans are performed periodically after initial use of the battery in the vehicle.
