Method and device for identifying kinetic parameters of electrochemical cell model by signal synchronization

Abstract

The invention relates to a computer-implemented method for determining or adapting at least one parameter of a battery model for at least one battery cell of a device battery of a battery system over a predetermined period of time on the basis of a time sequence of voltage and current signals of a cell voltage, the method comprises the following steps: providing a time sequence of voltage and current signals by sampling the sampled values of the cell voltage and the battery current at the same sampling rate; determining one or more signal time sequence segments according to the synchronization characteristics; performing a cross-correlation method on one or more signal time series segments of the voltage and current signal time series so as to obtain a signal shift between the voltage and current signal time series for each of the signal time series segments; synchronizing the voltage signal time sequence with respect to the current signal time sequence in one or more signal time sequence segments, respectively, depending on the signal displacement; at least one parameter of the battery model is determined or adapted as a function of the time series of voltage and current signals synchronized with each other.

Description

Technical Field

[0001] The invention relates to an electrochemical battery model for monitoring a battery and in particular for determining an aging state or for detecting anomalies. Background Art

[0002] For monitoring battery cells, the battery cells can be modeled with the aid of an electrochemical battery model that represents a system of nonlinear or differential equations. Such a battery model can be parameterized with the aid of equilibrium and kinetic parameters that describe the behavior of the battery cell and, in particular, the relationship between the cell current and the cell voltage. Since the equilibrium parameters and kinetic parameters depend on the aging state of the battery cell, the battery model must be re-parameterized at regular time intervals.

[0003] In particular, for adapting or determining kinetic parameters characterizing the dynamic behavior of a battery cell, time series signals of the cell current and the cell voltage are required with a relatively high resolution (e.g. 10 Hz). These time series signals are detected during system excitation, such as by charge and discharge current pulses, in order to obtain adequate excitation of the battery cell. Due to the high resolution and the required pulsed (high gradient) system excitation, it is necessary that the current and voltage signals are synchronized with each other, so that the changes in the curve of the cell current are synchronized in time with the corresponding curve of the cell voltage.

[0004] However, in practice, the curves showing the current and the cell voltage are often not synchronized with each other and may be shifted by one or two sampling periods. In addition, the temporal displacement between the cell current and the cell voltage is not constant over time and may change during the recording of the time series signal. Summary of the invention

[0005] According to the invention, a method for determining or adapting parameters of an electrochemical battery model for at least one battery cell of a device battery by means of synchronization of current and voltage signal time series is provided as claimed in claim 1 , and a corresponding device as claimed in the independent claim.

[0006] Further refinements are described in the dependent claims.

[0007] According to a first aspect, a computer-implemented method is provided for determining or adapting at least one parameter of a battery model for at least one battery cell of a device battery based on a voltage signal time series and a current signal time series of the cell voltage within a predetermined time period, the method comprising the following steps:

[0008] - Providing a voltage signal time series and a current signal time series by sampling the sampled values ​​of the single cell voltage and the battery current at the same sampling rate;

[0009] - determining one or more signal time series segments based on synchronization characteristics;

[0010] - performing a cross-correlation method on one or more signal time series segments of the voltage signal time series and the current signal time series, so as to obtain a signal shift between the voltage signal time series and the current signal time series for each of the signal time series segments;

[0011] - synchronizing the voltage signal time series with respect to the current signal time series in the one or more signal time series sections respectively according to the signal displacement;

[0012] - determining or adapting at least one parameter of the battery model based on the voltage signal time series and the current signal time series which are synchronized with each other.

[0013] As described at the beginning, in order to determine the kinetic parameters of the electrochemical cell model, it is necessary to provide time series signals of the current and voltage of the battery cell. These time series signals must be provided with high temporal resolution, i.e., at a frequency between 1 and 100 Hz. In addition, it is necessary that the current and voltage time series signals are synchronized with each other in time.

[0014] The method described below makes it possible to identify a shift between the current and voltage signal time series and to compensate for the shift accordingly. Since the cell current corresponds to the battery current in a device battery with a plurality of battery cells connected in series, it may be expedient to shift the cell voltage signal accordingly in the event of a shift between the current and voltage signal time series. One or more sampling periods are shifted forward or backward in time in sections, which are described by the signal shift.

[0015] It can be provided that, before synchronizing one or more signal time series segments of the voltage signal time series and the current signal time series, synchronization of the entire voltage signal time series and the current signal time series is performed, for example, by means of a cross-correlation method. In order to be able to achieve a global signal shift between the signal time series, a cross-correlation algorithm can be applied, which correlates the considered signal time series with each other and determines a dominant signal shift between the current signal time series and the voltage signal time series, which can, for example, include time series with a duration between 1 and 6 hours. The dominant signal shift is compensated, and the resulting undefined sampling value pairs at the beginning or end of the signal time series to be shifted can either be deleted or determined and filled by a suitable extrapolation method, in particular by means of a nearest neighbor method.

[0016] However, when detecting the time series signals, a local signal displacement occurs, which cannot be compensated by a global signal displacement. For this purpose, it is provided that signal time series segments in which the voltage signal time series and the current signal time series must be shifted relative to each other are identified, and the signal time series segments are correspondingly synchronized with each other by determining the signal displacement within the relevant signal time series segments. Therefore, it is first provided that signal time series segments that are asynchronous with respect to each other are identified. This can be done, for example, based on a synchronization feature, which evaluates the difference between the relative change between two temporally consecutive sampling values ​​of the voltage time series signal and the current time series signal. This is based on the fact that the current and voltage signals must be synchronized with each other. The synchronization feature for a battery cell can be defined as follows:

[0017]

[0018] in

[0019] Δi j =i j -i j-1 ;j∈2,3,4,…,n;Δi1=0

[0020] Δv j =v j -v j-1 ; j∈2, 3, 4,..., n; Δv1=0.

[0021] It can be provided that one or more signal time series segments are defined by a start index and an end index, wherein the start index indicates a sampling time point that is earlier in time than the end index, wherein the signal time series segment is determined based on the start index and the end index, in such a way that an index value of the signal time series is determined, in the case of which the corresponding characteristic value of the synchronization feature is numerically above a predetermined characteristic threshold, and in such a way that the start index and the end index are determined for each signal time series segment in the signal time series segment based on one of the specific index values ​​as the start index and a specific index value that follows in time as the end index, wherein the end index primarily has a characteristic value with a sign different from the characteristic value of the synchronization feature at the start index.

[0022] Thus, the difference or the difference between the relative change of the voltage time series signal and the relative change of the current time series signal can be evaluated. If a change in value above a predetermined characteristic threshold is detected, a potential signal segment shift is detected. This is performed specifically by a threshold comparison with a predetermined characteristic threshold, which can be, for example, between 0.02 and 0.1, preferably at 0.06.

[0023] The above synchronization features can be used to find possible start indexes and end indexes (sampling time points), which can represent the start and end of the corresponding shifted signal time series segment. Finally, the signal time series segment to be identified between the corresponding start index and the corresponding end index is identified, and the characteristic value of the signal time series segment exceeds the predetermined characteristic threshold value according to the numerical value and has a sign change. The determination of the sign change is used to determine the boundary of the signal time series segment.

[0024] For this purpose, for example, all index values ​​of the considered signal time series whose characteristic values ​​are greater than a predefined characteristic threshold value can be written into the index vector Hence, all (specific) index values ​​thus found are adopted as potential candidates for the start index or end index of a signal time series segment within the entire signal time series.

[0025] The start index can be determined when the start index is the first index value in the list of specific index values ​​or as a specific index value after the last determined end index, at which a sign change of the characteristic value occurs relative to the characteristic value of the end index. In this case, the search is performed starting from the last index value of the index vector forward, i.e. in the direction of decreasing index values. For example, a sign change of the characteristic value can occur when a transition occurs between a charging behavior and a discharging process.

[0026] After each starting index is found, move backwards in the direction of increasing index values ​​from the index vector The index value for which a change of sign of the eigenvalue occurs is searched in .

[0027] Therefore, the index value with the sign change of the feature value represents the start index or the end index for the signal time series segment.

[0028] It can be provided that overlapping signal time series segments are connected to one another. This can be done by following the start index value j with the end index s，i -j e，i-1 ≤ 1 (for i = 2...n) to identify overlapping windows, where j s The starting index value and j corresponding to the signal time series segment e Corresponds to the end index value. If the condition is met, the start index j is removed accordingly s，i and end index j e，i-1 The index value of .

[0029] It may be provided that for a signal time series segment of the determined signal time series segments which has a number of sampling values ​​which is less than a predefined minimum number, the number of sampling values ​​of the relevant signal time series segment is increased.

[0030] Since the cross-correlation analysis does not work reliably in the case of too small a number of sampling values, a very small time series segment of less than a predefined number of sampling values, for example five sampling values, can therefore be extended.

[0031] A time series segment that is too small can be correspondingly enlarged by shifting the index values ​​of the start index and the end index forward or backward to a small time segment in order to give the time series segment a sufficient size.

[0032] j s，i =j s，i -[0.5(j e，i -j s，i -w]

[0033] j e，i =j e，i +[0.5(j e，i -j s，i -w]

[0034] where w corresponds to the minimum size of a signal time series segment.

[0035] The above formula enables the start and end indices to be shifted by integer index values ​​and causes odd intervals of sample values ​​to remain odd.

[0036] After the size of the too small time series segments has been adapted, the merging of adjacent and overlapping time series segments can be performed again.

[0037] Subsequently, the voltage and current signals can be correlated with each other in the signal time series segments. The correlation is performed by applying a cross-correlation algorithm to the corresponding signal time series intervals, so that the value of the signal segment displacement is determined so that the voltage signal time series can be shifted by said value. This determination is performed according to the determination of the maximum value of the correlation function and corresponds to its displacement relative to the zero position.

[0038] Gaps in sample value pairs resulting from the displacement of the signal time series segment are overwritten by NaN values. At the correspondingly different ends of the time series segment, the sample values ​​are shifted out of the time series segment and the corresponding sample values ​​of the signal time series before or after the signal time series segment are overwritten there. After local synchronization of all time series segments, the NaN samples are filled with interpolated sample values, in particular by means of linear interpolation between the previous and subsequent known sample values.

[0039] In this way, it is possible to restore synchronization by global and local synchronization of the two signal time series even if changes in the local variations of the signal displacements of the current and voltage signal time series occur within the signal time series.

[0040] The synchronized current and voltage signal time series can be used to parameterize an electrochemical cell model, which is based on, for example, a system of nonlinear equations or a system of differential equations with multiple nonlinear differential equations. The current and voltage signal time series enables the current battery state to be modeled by means of a time integration method. Such an electrochemical cell model is known, for example, from publications US 2020 / 150,185A, WO2022017984A1, EP3919925A1, DE102020206915B3 and US20210373082A1. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The following is a more detailed description of the implementation method based on the attached drawings.

[0042] Figure 1 A schematic diagram showing a system for determining kinetic parameters of a device battery based on current and voltage signal time series; and

[0043] Figure 2 A flow chart illustrating a method for synchronizing current and voltage signal time series to determine kinetic parameters of an electrochemical cell model is shown.

[0044] Figure 3 3a and 3b show the curves of the time series of current and voltage signals and the illustration of the synchronization characteristics for determining local time series sections with locally different displacements. DETAILED DESCRIPTION

[0045] Figure 1 The schematic diagram shows a battery system 1 having a plurality of battery cells 2 of an equipment battery, which are connected in series in order to provide a desired battery voltage.

[0046] A battery management system 3 is provided, which monitors the cell voltage and cell current / battery current of a plurality of battery cells 2. A mathematical model is implemented in the battery management system 3 in the form of an electrochemical battery model 31, which can perform aging states and abnormality identification of device batteries of the battery system 1 based on time series signals of current and cell voltage.

[0047] The electrochemical battery model 31 makes it possible to determine the internal electrochemical state, in particular the equilibrium state and the kinetic state, based on the current time series signal and the voltage time series signal, from which the corresponding aging state of the battery system 1 can be derived in a manner known per se, or based on the internal electrochemical state, functional abnormalities of the battery cells 2 of the entire battery system 1 can be identified with the aid of a suitable abnormality identification model.

[0048] In order to monitor the battery cells 2, a plurality of battery cells 2 can each be equipped with a voltage sensor 4 in order to detect the respective cell voltage and provide the respective cell voltage in the form of sampled values ​​in the battery management system 3. In addition, the battery current flowing through all battery cells can be detected by means of a current sensor 5 and provided in the form of sampled values ​​in the battery management system 3.

[0049] The sampling rate for detecting the cell voltage and the battery current may be between 1 and 100 Hz. Preferably, the sampling is performed at 10 Hz or in the range of 10 Hz. Due to the aging of the battery cells 2 of the battery system 1, the parameters of the electrochemical battery model need to be adapted regularly in order to reflect changes in the behavior of the device battery due to aging in the electrochemical battery model.

[0050] It has been shown in practice that the detection of the time series signals of the cell voltage and the battery current can be shifted relative to each other by one or more sampling values. In addition, different local shifts can be performed within the time series segment.

[0051] according to Figure 2 A method is now described in which time series signals shifted from each other are synchronized with each other. Since the battery current flows through all battery cells 2, synchronization is preferably performed by shifting and adapting the voltage time series signal of the cell voltage according to the following method.

[0052] To this end, in step S1, for each of the battery cells 2, the time series signals of the cell voltage and the battery current are detected at the same sampling rate within a predetermined period of 1 to 6 hours. Figure 3 A shows the time series of the current and voltage signals over a duration of 5 seconds. The displacement of the curves of the current and voltage is recognized, and in the case of a battery the edges of the curves must be synchronized with one another.

[0053] The following steps are now carried out individually for each of the battery cells 2 .

[0054] In step S2, a global synchronization of the respective voltage and current signal time series is now performed by means of a cross-correlation method. The cross-correlation method enables the determination of a global signal shift in the form of a plurality of sample values, with which the respective voltage signal time series is shifted relative to the current signal time series.

[0055]

[0056] where i corresponds to the current sampling value, and v corresponds to the voltage sampling value, and for j=2...n, Δi t =i j -i j-1 and Δv j =v j -v j-1 , where n corresponds to the total number of sampled values ​​of the signal time series under consideration.

[0057] In step S3, according to the determined global signal displacement, the voltage signal time series is shifted by a corresponding number of X sampling values ​​to establish synchronization with the current time series signal. Incomplete sampling value pairs obtained at the beginning or end of the time period are filled with reasonable values ​​or removed from the signal time series by appropriate extrapolation methods.

[0058] In order to determine the local signal displacement of the voltage signal time series and the current signal time series that deviates from the global synchronization, the characteristic value of the synchronization feature can be first determined for each of the sampled values ​​in step S4. The synchronization feature is defined as follows:

[0059]

[0060] And if, depending on the dynamics of the cell current, the cell voltage does not directly follow the current pulse or even precede it, then a deviation is indicated. Figure 3 In b, the curve of the synchronization characteristic s is shown for an exemplary signal time series of current and voltage. If a phase shift is present, deviations of the characteristic values ​​are detected, in particular in the dynamic phase.

[0061] In step S5, the characteristic values ​​of the synchronization characteristics can be compared with a threshold value in each case, which can be predetermined, for example, in the range between 0.02 and 0.1, in particular 0.06. The elements of the index vector corresponds to the index value for which the feature value has numerically exceeded the feature threshold, where i corresponds to the element index of the index vector, where j = 1 ... m, where m is the index vector The number of index values ​​in .

[0062] In step S6, now starting from the first index value, the time series segment is determined from a series of specific index values ​​of the index vector. After the first index value is used as the starting index value, the end index value for which the sign change of the eigenvalue of the synchronization feature occurs with respect to the sign of the eigenvalue of the synchronization feature of the last determined starting index is determined as the next index value of the index vector. The next specific index value after the end index value and for which there is a sign change for the sign of the eigenvalue at the last determined ending index is used as the next starting index value. After the starting index, as described above, the end index for which the sign change of the eigenvalue of the synchronization feature of the corresponding starting index occurs is determined as the next index value of the index vector. This is performed for the entire signal time series to be analyzed of the single cell voltage and the single cell current.

[0063] Therefore, a signal time series segment is defined using a start index value and an end index value.

[0064] From the list of start index values ​​and end index values, it may be possible that the time series segments are adjacent to or overlap each other. If it is determined that the end index value of a time series segment is directly before the start index value of the next time series segment or the end index value of a time series segment is after the start index value of the next time series segment, then in step S7, the time series segments can be connected to each other.

[0065] If this is the case, the relevant start index value and end index value are removed so that the start index value of the preceding signal time series segment and the end index value of the following signal time series segment constitute a connected time series segment.

[0066] Furthermore, in step S8, a time series segment containing a small number of sampling values ​​(eg, less than five sampling values) may be enlarged, in particular by shifting the start index value and the end index value of the relevant time series segment forward or backward, while maintaining an odd number of sampling value pairs.

[0067] If it is determined in step S9 that at least one of the time series segments has been enlarged (option: yes), the method continues with step S7. Otherwise (option: no), the method continues with step S10.

[0068] In step S10, for each identified time series segment determined by the start index value and the end index value, a cross-correlation method may be performed to determine a signal shift value indicating the number of sampling values ​​by which the associated voltage time series signal should be shifted.

[0069] In step S11, the sampled values ​​are moved forward or backward in time from the time series segment according to the shift value, and the corresponding sampled values ​​of the adjacent time series segment are overwritten. The gaps in the sampled value pairs due to the signal shift are first filled with NaN values.

[0070] If a corresponding signal displacement value is determined for each of the time series segments by means of a cross-correlation method and applied, then corresponding sampling values ​​are generated by interpolation at sampling time points at which NaN values ​​have been entered. The interpolation can be performed by linear interpolation.

[0071] In step S12 , the synchronized signal time series of the cell voltage and the cell current generated in this way can be used to determine or adapt parameters of the electrochemical battery model.

[0072] The electrochemical cell model can be used in a manner known per se to determine the state of aging and / or to identify anomalies.

Claims

1. A computer-implemented method for determining or adapting at least one parameter of a battery model (31) for at least one battery cell (2) of a device battery of a battery system (1) based on a voltage signal time series and a current signal time series of the cell voltage within a predetermined time period, the method comprising the following steps: - Providing (S1) a voltage signal time series and a current signal time series by sampling the sampled values ​​of the single cell voltage and the battery current at the same sampling rate; - determining (S6) one or more signal time series segments based on the synchronization characteristics; - performing (S10) a cross-correlation method on one or more signal time series segments of the voltage signal time series and the current signal time series, so as to obtain a signal displacement between the voltage signal time series and the current signal time series for each of the signal time series segments; - synchronizing the voltage signal time series with respect to the current signal time series in the one or more signal time series segments according to the signal displacement (S11); - Determining or adapting (S12) at least one parameter of the battery model (31) according to the voltage signal time series and the current signal time series which are synchronized with each other.

2. The method according to claim 1, wherein before synchronizing one or more signal time series segments of the voltage signal time series and the current signal time series, synchronization of the entire voltage signal time series and the current signal time series is performed (S3) by means of a cross-correlation method. 3 . The method according to claim 1 , wherein the synchronization feature specifies, for each sampling time point, a difference, in particular a difference, between relative changes between two temporally consecutive sampling values ​​of the voltage signal time series and the current signal time series.

4. The method according to any one of claims 1 to 3, wherein the one or more signal time series segments are defined by a start index and an end index, wherein the start index specifies a sampling time point earlier in time than the end index, The signal time series segment is determined based on the start index and the end index, in that an index value of the signal time series is determined, at which index value the corresponding characteristic value of the synchronization feature is numerically above a predetermined characteristic threshold, and in that the start index and the end index are determined for each signal time series segment in the signal time series segment based on one of the specific index values ​​as the start index and a specific index value that follows in time as the end index, the end index primarily having a characteristic value with a different sign than the characteristic value of the synchronization feature at the start index.

5. The method according to claim 4, wherein for a signal time series segment in the determined signal time series segments having a number of sampling values ​​less than a predetermined minimum number, the number of sampling values ​​of the relevant signal time series segment is increased (S8). 6 . The method according to claim 4 , wherein overlapping or directly adjacent signal time series segments are connected to one another ( S7 ). 7 . The method according to claim 1 , wherein gaps in the sampled values ​​resulting from the displacement of the signal time series segments are filled by interpolation of the sampled values, in particular by linear interpolation.

8. An apparatus for performing the method according to any one of claims 1 to 7.

9. A computer program product comprising instructions which, when executed by at least one data processing device, cause the data processing device to perform the steps of the method according to any one of claims 1 to 7.

10. A machine-readable storage medium comprising instructions which, when executed by at least one data processing device, cause the data processing device to perform the steps of the method according to any one of claims 1 to 7.

Images