DE102015011745B3 - Charge state determination in a motor vehicle battery - Google Patents

Abstract

The invention relates to a method for determining a state of charge (8) of an electric battery (2) in a motor vehicle (1), wherein in successive measuring cycles (k) a respective measured value (5) of at least one operating variable (UBATT, IBATT, T) of the Battery (2) is measured and calculated by a Kalman filter device (9) in a correction step (29) a current state vector from a previous state vector and a current covariance of an estimation error from a previous covariance of the estimation error and in a prediction step (30) a predicted state vector (33) is calculated from the current state vector, and a predicted covariance is calculated from the current covariance, and in an output step, an estimated state of charge value (SOC) included in the new predicted state vector (33) is output as the state of charge (8). The estimation of the state of charge value (SOC) should be robust. The invention provides that prior to the correction step (29) in the preceding state vector or before the prediction step (30) in the current state vector, the state of charge value is replaced by an estimated value determined by an estimation device (10) independent of the Kalman filter device (9) ) is determined for the state of charge.

Description

The invention relates to a method for determining a state of charge of an electric battery in a motor vehicle. In the method, a respective measured value of at least one operating variable of the battery is measured in successive measuring cycles and a state vector is predicted by a Kalman filter device, which contains as an input vector an estimated state of charge value, which is output as the state of charge. The invention also includes an apparatus for carrying out the method. Finally, the invention includes a motor vehicle having the device.

The exact determination of the state of charge of a motor vehicle battery, in particular a lithium-ion battery, presents a challenge for the signal processing. The requirements for the accuracy of the determined state of charge are of great importance, in particular in the automotive field of application, in order to ensure safe and secure operation in a motor vehicle efficient operation of the electrochemical energy store, that is the battery, to be able to guarantee.

From the DE 10 2013 018 405 A1 For example, a method for determining an estimation parameter characterizing the state of a battery is known in which a calculation result of a support vector machine uses a Kalman filter as the initial start variable.

From the publication by Zhang and Lee (2011, A review on prognostics and health monitoring of Li-ion battery, Journal of power sources, 196 (15), pp. 6007-6014) there is given an overview of common methods, a state of charge of a To determine battery. Among other things, the use of an "extended Kalman filter" is addressed.

Already known methods based on an equivalent electrical circuit diagram, such as the Kalman filter, the extended Kalman filter or the condition observer, are reaching their limits here. The reason for this are two important points that play a role, in particular in lithium iron phosphate technology: on the one hand the very flat plateau of the quiescent voltage characteristic in the middle state of charge and on the other hand the accuracy of the electrical equivalent circuit with respect to the network parameters used to calculate the quiescent voltage be based on the terminal voltage. Thus, in the case of the rest-voltage-based methods, the state of charge determination in the middle range with accuracies in the single-digit percentage range can only be realized with very great metrological effort.

The invention has for its object to determine a state of charge of a motor vehicle battery.

The object is solved by the subject matters of the independent claims. Advantageous developments of the invention are given by the features of the dependent claims.

The invention provides a method for determining a state of charge of an electric battery in a motor vehicle. Under battery is to be understood in the context of the invention, an electrochemical accumulator, in particular a lithium-ion battery. In the method, a respective measured value of at least one operating variable of the battery is measured in successive measuring cycles. By a Kalman filter device, the following, known per se steps are performed. In an update step or correction step, a current state vector from a previous state vector each containing a state of charge load state value and a current covariance of an estimation error of the Kalman filter device are calculated from a previous covariance of the estimation error. The current state vector can be determined in a manner known per se from a predicted state vector of the immediately preceding measurement cycle, which thus represents the preceding state vector. In the same way, the current covariance can be determined from a predicted covariance of the immediately preceding measurement cycle, which thus represents the preceding covariance.

From the current state vector, a new predicted state vector is calculated in a prediction step, and a new predicted covariance is calculated from the current covariance. The calculations necessary for the correction step and the prediction step are known per se from the theory for the Kalman filter. The new predicted state vector contains an estimated state of charge value. This estimated state of charge value is output in an output step as the current state of charge.

In order to compensate for the described shortcomings or disadvantages of the Kalman filter at a flat plateau of the rest voltage, the invention provides that before the correction step ( 29 ) in the previous state vector or before the prediction step in the current state vector, the state of charge value is replaced by an estimated value determined by an estimator independent of the Kalman filter device. In other words, the original current state vector is not used to compute the predicted state vector, but a more manipulated actual state A state vector in which a vector entry, namely the state of charge value, has been replaced by an estimated value determined by means of the independent state-of-charge estimator.

The advantage of the invention is that the state of charge state value of the current state vector and the corresponding state state vector state of charge value are decoupled in that two different estimators are used, firstly the Kalman filter device and the second estimation device independent of the Kalman filter device become. As a result, the covariance of the estimation error can be reduced or stabilized, in particular in a flat plateau of the quiescent voltage characteristic.

The invention also includes optional developments, by the characteristics of which additional benefits.

According to a development, the at least one operating variable that is used for the estimation comprises a terminal voltage of the battery and / or a battery current of the battery and / or a temperature of the battery. These operating variables are advantageously measurable from the outside, that is they represent suitable observation variables. The internal operating state, namely in particular the state of charge of the battery, is then determined from these operating variables by means of the Kalman filter device and the estimating device.

According to a development, the estimation device additionally determines a variance value of the estimation variance. Before the correction step in a covariance matrix describing the preceding covariance or before the prediction step, this variance value is stored as a matrix entry in a covariance matrix describing the current covariance of the estimation error. In other words, before the correction step or before the prediction step, not only is the charge state value overwritten or replaced, but also the value for the covariance. This results in a more accurate fit of the Kalman gain or Kalman gain, which is then based on the variance of the state of charge estimate as determined by the independent estimator.

In order to achieve a suitable alternative estimation of the state of charge value by means of the estimation device, a respective time series for the at least one operating variable is formed by the estimation device on the basis of the measured values of a plurality of measurement cycles. Thus, a time series for the terminal voltage and / or a time series for the battery current can be formed. By means of a frequency analysis, for example a Fourier transformation, a frequency-dependent impedance of the battery is determined from the at least one time series. This results in the advantage that on the one hand by means of the Kalman filter means a charge state estimation on the basis of a signal in the time domain and by means of the independent estimator an estimate of the state of charge is determined on the basis of a signal in the frequency domain. This results in a particularly robust estimation of the state of charge.

According to a further development, one of a plurality of predetermined classes is selected by the estimating device by means of a support vector machine by classifying the measured values or values derived therefrom, each class representing a state of charge. The derived values can, for. B. include the described frequency-dependent impedance of the battery. The selected class then determines the estimated value of the estimator for the state of charge, which is then entered into the current state vector. This development has the advantage that it does not use an integrating state estimation, as is necessary in the case of the Kalman filter device, which is based on a state space model which provides ampere-hour integration. By contrast, by means of the support vector machine, the classification of the measured values provides an estimate of the state of charge directly by determining or selecting the current class.

A further development takes into account that the Kalman filter device on the one hand and the estimating device independent of it on the other hand can be differently accurate for different operating situations. In this development, the state of charge value is only replaced by the estimated value of the estimating device if, in at least one preceding measuring cycle, the determined state of charge lies in a predetermined state of charge state. In this way, for example, the described flat plateau of the quiescent voltage characteristic can be delimited, for which then the estimator is used while the state of charge of the battery outside the state of charge, ie fully charged or nearly fully charged battery on the one hand and almost discharged or discharged battery on the other hand exclusively uses the Kalman filter device can be.

A further development in this connection provides for making the selection between the Kalman filter device and the independent estimator dependent on the current estimation quality. In this development, the state of charge state is only replaced by the estimated value of the estimator if the preceding or the current covariance violates a predetermined quality criterion. This development has the advantage on that regardless of the current state of charge, the selection takes place in dependence on the current estimation quality. The quality criterion may, for example, provide for a limit value or a value interval for the covariance. Below the limit value or within the value interval, only the Kalman filter device is then used, while above the limit value or outside of the value interval the additional estimated value is then determined in the manner described by means of the estimation device, which is then inserted into the current state vector.

The invention also includes a device for a motor vehicle, by means of which the method according to the invention can be carried out. The device can be configured, for example, as a control device, in particular as a battery control device. The device has a measuring input for receiving respective measured values of at least one operating variable of the battery. Furthermore, a computing device is provided which can be realized, for example, by means of a microcontroller or a microprocessor. The computing device comprises a Kalman filter device for estimating a state of charge of a battery of the motor vehicle as well as an estimation device independent of the Kalman filter device for generating an estimate for the state of charge. The Kalman filter device and the estimation device can be realized, for example, on the basis of one or more program modules for the computing device. The device according to the invention is designed to carry out an embodiment of the method according to the invention.

Finally, the invention also includes a motor vehicle. The motor vehicle has a battery, in particular a lithium-ion battery. Furthermore, a measuring device is provided for measuring measured values of at least one operating variable of the battery. This is in particular a terminal voltage of the battery and / or a battery current of the battery and / or a temperature of the battery. Finally, an embodiment of the device according to the invention is provided in the motor vehicle, which is coupled to the measuring device.

The motor vehicle according to the invention is preferably designed as a motor vehicle, in particular as a passenger car.

In the following an embodiment of the invention is described. Showing:

1 a schematic representation of an embodiment of the motor vehicle according to the invention,

2 a diagram with rest voltage characteristics,

3 a battery model,

4 a schematic representation of a Kalman filter device,

5 a diagram with frequency-dependent impedances,

6 a sketch to illustrate a support vector machine,

7 a sketch to illustrate a classification, and

8th a schematic representation of an embodiment of the device according to the invention.

The exemplary embodiment explained below is a preferred embodiment of the invention. In the exemplary embodiment, the described components of the embodiment each represent individual features of the invention that are to be considered independently of one another, which also each independently further develop the invention and thus also individually or in a different combination than the one shown as part of the invention. Furthermore, the described embodiment can also be supplemented by further features of the invention already described.

In the figures, functionally identical elements are each provided with the same reference numerals.

1 shows a motor vehicle 1 which may be, for example, a motor vehicle, in particular a passenger car. The car 1 has a battery in the example shown 2 , a measuring device 3 and a device 4 on. The battery 2 For example, a traction battery for an electric drive motor (not shown) of the motor vehicle 1 be. The battery 2 may be, for example, a lithium-ion battery. In particular, it may be based on lithium iron phosphate technology. By means of the device 4 can be a state of charge SOC of the battery 2 be determined. The measuring equipment covers this 3 readings 5 of operating parameters or operating parameters of the battery 2 , The measuring device 3 For example, a temperature sensor 6 include a temperature T of the battery 2 detected. Furthermore, the measuring device 3 be designed in known manner, a battery current I _Batt and a terminal voltage U _Batt of terminals 7 the battery 2 to capture. The device 4 can be configured for example as a control unit. The device 4 determined on the basis of the measured values 5 an estimated state of charge value 8th the state of charge SOC.

2 illustrates what problem can arise in estimating the state of charge SOC. In 2 shows over the possible values of the state of charge SOC in percent (%) each resulting in-battery residual voltage U _OCV in volts for a battery cell of the battery 2 , This results in a charging curve 11 and a discharge curve 12 , each representing a quiescent voltage characteristic. In a state of charge area 13 Both have a quiescent voltage characteristic 11 . 12 a flat plateau 14 on. Here even a small estimation error of the rest voltage U _OCV results in a large change in the estimated state of charge SOC. The estimation errors may be due, in particular, to the fact that the quiescent voltage U _OCV can not be measured directly.

This is in 3 again using a battery model 15 the battery 2 illustrated. The by the measuring device 3 measured terminal voltage U _Batt is composed of the rest voltage U _OCV and via the internal resistance R _i and an RC element of the model 15 decreasing voltages U _Ri and U _RC . In this case, the current temperature T has an influence.

The Kalman filter device 9 is for generation of an estimated state of charge value 8th outside the state of charge 13 particularly suitable. To the described weakness of the Kalman filter device 9 in the charge state area 13 is the additional estimator 10 provided.

4 shows the Kalman filter device 9 in detail. Through the Kalman filter device 9 a Kalman filter is realized by which a state vector x _{k is} estimated. The state vector x _k is updated in each case for individual iterations or measurement cycles k. The state vector x _k comprises state of charge value SOC _k for the current state of charge SOC, as well as a current estimated value U _{RC, k} for the battery-internal voltage U _RC dropping across the RC element (see FIG 3 ):

As input vector u _k , the current current measurement for the battery current I _Batt can be used: u _k = I _{batt, k} .

For example, the determined terminal voltage U _Batt can be used as the system output variable y _k .

In the 4 shown filter components based on the battery model 15 from 3 and represent, in a manner known per se, the input matrix b, the passage matrix d, the output matrix c, the Kalman gain or the Kalman gain k _k and the system matrix A. The Kalman filter device 9 serves the process and measurement noise suppression and responds to stochastic interference, with appropriate system knowledge according to the battery model 15 the system behavior of the battery 2 to be able to predict correctly. Similar to the state observer, the Kalman filter uses feedback as described in Equation (2) below to reduce the error between the model's output and the system output y _k with the Kalman gain k _k . By balancing the difference, the state of the model also approaches that of the system. In contrast to the state observer, however, the Kalman filter works in two steps according to the principle of mutual recursion.

In this update step or correction step, the Kalman gain k _{k is also} calculated. This is in contrast to the gain of the state observer not a fixed pole, but is recalculated with each time step or measurement cycle k over stochastic approaches with the covariance of the estimation error P _k according to the following equation (3) and the covariance of the system noise R _k (fixed) or corrected, as described in the following equation (1), in order to ensure an optimal excitation of the difference with high dynamics can. In this way, the measurement and system noise can be optimally suppressed. k _k = P _k-1 ^T c (cP c _k-1 ^T + R _k) ^-1 (1) x ^ _k = x ^ _k-1 + k _k (y _k - cx ^ _k-1) (2) P _k = (I - k _k c) P _k-1 (3)

In a prediction step, the Kalman filter calculates the future system state as described by the following equation (4). The correction of the covariance P _{k + 1 of} the estimation error for the next calculation of the Kalman gain describes the following equation (5) on the covariance of the measurement noise Q _k (fixed). x ^ _{k + 1} = Ax ^ _k + BU _k (4) P _{k + 1} = AP _k A ^T + Q _k (5)

The quality of the prediction in the form of the covariance of the estimation error and thus an essential property of the Kalman filter is thus recalculated for the next correction.

The following is based on 5 and 6 the estimator 10 describe. It can have a support vector machine SVM. A support vector machine is not a machine in the traditional sense but is a mathematical method for classifying sets and classifying elements into a set.

The treasury 10 based on the in 5 illustrated fact that impedance loci or short loci 16 (Frequency response locus) of an impedance Z of a lithium-ion battery from the current state of charge SOC of the battery 2 is dependent. In particular, this applies to a lithium iron phosphate battery. 5 represents the impedance loci 16 above the real part Re and the imaginary part Im for several different exemplary charge states SOC1, SOC2, SOC3, ..., SOCn. Each locus 16 represents the course of the impedance Z for different frequencies.

In 6 and 7 is the assignment of a locus 16 to a state of charge SOC as performed by the SVM. In the case of the SVM, the individual states of charge are each combined in one class, which are referred to here as state of charge class c ₁ , c ₂ ,..., C _n .

The state of charge classes c ₁ , c ₂ , ..., c _n become, as in 6 illustrated by an example, demarcated by SVM in the form that each charge state class is represented by a hyperplane computed by the SVM 21 is clearly separated to a nearest other state of charge class. 6 illustrates how, based on training values for a first state of charge + and a second state of charge * parameters of the hyperplane 21 during training or configuration of the SVM. The hyperplane 21 separates the two training sets +, *. With an observation value for an impedance Z, this observation value becomes one side of the hyperplane 21 classified.

In order to separate all charge state classes c ₁ , c ₂ ,..., C _n from each other, each state of charge class must be at its closest to a hyperplane 21 be spatially separated. Thus, in n classes, n - 1 hyperplanes result 21 which subdivide, that is classify, the classes. The resulting set of hyperplanes thus represents the decision basis of the SVM with regard to a measured impedance locus to be classified 16 represents.

The entire classification is through 7 illustrated. An impedance Z detected during operation can now be classified by means of the SVM. However, since an SVM, by means of its calculated hyperplanes, is only able to make a binary decision (yes, no) for each individual hyperplane, a method is needed which evaluates the decisions of the individual hyperplanes and leads to a clear classification of the impedance Z. By mapping the set of all charge state classes and their power sets in a binary search tree, this problem of the evaluation can be solved efficiently. In the search tree according to 7 represents the root 23 the set of all state of charge classes. The two following nodes 25 of the binary search tree represent the two power sets of the superset separated by the median of the hyperplane. The median is determined by the fact that with respect to the hyperplane of the superset (that is, first of the root 23 ) an equal number of hyperplanes is arranged on both sides of and beyond the hyperplane of the superset. This division of the power set can now be continued in subsequent steps until the search tree is completely filled up and the individual sheets correspond to the respective charge state classes c ₁ , c ₂ ,..., C _n . Ideally log ₂ (n) follow-up steps result in n state of charge classes.

This results in a search algorithm according to the principle "divide and rule", which is now able to unambiguously assign a measured impedance Z to a state of charge class c ₁ , c ₂ ,..., C _n . Not even all hyperplanes need to be evaluated in this procedure.

If a measured impedance Z is to be classified, the SVM uses the relevant hyperplane and the median of the hyperplanes of the respective power set (root / node) to decide whether the impedance belongs to the power set above or below the hyperplane. Every edge 24 in the search space this represents a decision of the respective hyperplane. In the classification of several impedances Z can in this way a statistical distribution 22 the charge states in the form of a histogram with an estimate μ _SVM for the state of charge and a variance σ 2 / SVM to the expected value μ _SVM .

The use of SVM to estimate the state of charge is also disclosed in a paper by Jansen et al. (P. Jansen, D. Vergossen, D. Renner, W. John, J. Götze, "Impedance spectra classification for determining the state of charge on a lithium iron phosphate cell using a support vector machine", Advances in Radio Science, Volume 13, 2015, doi: 10.5194 / ars-13-1-2015, www.adv-radio-sci-net / 13/1/2015).

The impedance Z can be determined from the measured values 5 for the terminal voltage U _Batt and the battery current I _Batt . In 8th This is illustrated by how the estimator 10 a transformation device 17 comprising, by means of a Fourier transformation, a series of measurements 18 for the terminal voltage and a series of measurements 19 for the battery current each by means of a transformation 20 Frequency values or spectral values generated by means of which the impedance Z can be calculated in a known manner. The transformation 20 can be based on a fast Fourier transform FFT. The impedance Z and the determined battery temperature T can be provided as input parameters for the support vector machine SVM.

8th further illustrates how the Kalman filter device 9 and the estimator 10 can be combined.

For a hybrid state of charge determination, a coupling of the two methods, that is the Kalman filter device 9 and the estimator 10 so through the Kalman filter device 9 in the time domain and by the SVM in the frequency domain, the charge state determination is carried out. This results in synergistic effects. The coupling should not explicitly be implemented via a weighting function, but an interference in the mutual recursion of the Kalman filter over the obsolete state vector with the results from the measurement update and the evaluation of the measurement by the covariance matrix. In the form that the classification result of the SVM, ie the estimated value μ _SVM for the state of charge and the variance derived by means of the described determination method σ 2 / SVM of the classification error, overwrite the old values of the Kalman filter.

This is in 8th represented as by the measurement 27 the system output y _k with a measurement variance 28 is detected. In the already described correction step 29 For example, the Kalman gain k _{k is} executed according to the above equation (1) and the corrections or updates based on the above equations (2) and (3).

As in the presentation of 8th can be shown through a time update afterwards 31 the next measurement cycle begin, ie the index k now applies. In the time update 31 In the current state vector x _k, the vector entry for the state of charge SOC is replaced by the estimated value μ _SVM . In the covariance matrix P _k becomes the matrix entry 32 for the current covariance of the estimation error of the state of charge SOC replaced by the variance σ 2 / SVM, like the one through the estimator 10 has been determined. The replacement can also be in the correction step 29 in the preceding state vector x _k-1 and in the covariance matrix P _k-1 in the preceding covariance σ 2 / k-1 respectively.

After that, the prediction step 30 are performed according to equations (4) and (5). This results in a predicted state vector 33 of which the described estimated state of charge value 8th of the state of charge SOC can be read or removed.

In this way, the mutual recursion or iteration of the Kalman filter and the measurement may be affected so that in the subsequent prediction of the state vector by the Kalman filter based on the overwritten covariance matrix, the estimate of the Kalman filter is discarded.

This can optionally by a verification device 34 First, it is to be checked whether the Kalman filter is to use the prediction results for state of charge determination instead of the estimates of the estimator because of the quality of the covariance matrix P _k 10 ,

Overall, this allows information about the state of charge of the battery 2 be used in the time and in the frequency range for state of charge determination. By such an implementation, the advantages of both devices, that is the Kalman filter device 9 and the estimator 10 , for which cell chemistry lithium iron phosphate are linked together. The very good state of charge determination of the Kalman filter in the upper and lower state of charge, ie outside the plateau 14 of the charge state area 13 , as well as the first state of charge determination of SVM in the medium state of charge 13 By means of classification can also be used by the fact that the estimator 10 is only used for such states of charge that are within the state of charge 13 are located.

For frequency domain processing by SVM, battery impedance values of specific frequencies are identified based on consumer excitation in the vehicle electrical system. This can be done, for example, by the fast Fourier transformation FFT described.

The coupling of the Kalman filter device 9 and the estimator 10 takes place via the obsolete state vector x _k-1 in the time update 31 , For this purpose, the system state and the evaluation of the old measurement by the covariance matrix P _k-1 from the previous calculation step are used. For this purpose, the classification result of the SVM, ie the maintenance value of the state of charge μ _SVM and a variance derived from the underlying determination method σ 2 / SVM of the likely classification inaccuracy. The inadequately accurate elements of the matrices of the Kalman filter x _k-1 and P _k-1 , from the previous calculation step, will now be used with the new results of SVM (μ _SVM and σ 2 / SVM) overwritten during stay in the middle charge state area.

Overall, the example shows how the invention can provide a hybrid charge state determination of batteries by means of coupled information from the frequency range and the time range.

Claims (10)

Method for determining a state of charge ( 8th ) of an electric battery ( 2 ) in a motor vehicle ( 1 ), wherein in successive measuring cycles (k) in each case a respective measured value ( 5 ) at least one operating _variable (U _BATT , I _BATT , T) of the battery ( 2 ) and by a Kalman filter device ( 9 ): - in a correction step ( 29 ) a current state vector from a previous state vector, each containing a state of charge value, and a current covariance of an estimation error from a previous covariance of the estimation error are calculated and - in a prediction step ( 30 ) from the current state vector a predicted state vector ( 33 ) and a predicted covariance is calculated from the current covariance, and - in an output step, a predicted state vector in the new ( 33 ), the estimated state of charge value (SOC) as the state of charge ( 8th ), characterized in that before the correction step ( 29 ) in the preceding state vector or before the prediction step ( 30 ) in the current state vector, the respective state of charge value is replaced by an estimated value obtained by one of the Kalman filter devices ( 9 ) independent estimator ( 10 ) is determined for the state of charge. The method of claim 1, wherein the at least one operating _variable (U _BATT , I _BATT , T) _comprises a terminal voltage (U _BATT ) and / or a battery _current (I _BATT ) and / or a temperature (T). Method according to one of the preceding claims, wherein the estimation device ( 10 ) additionally a variance value of the estimated value is determined and before the correction step ( 29 ) in a covariance matrix describing the preceding covariance or before the prediction step ( 30 ) is stored in a covariance matrix describing the current covariance of the estimation error, the variance value as a matrix entry. Method according to one of the preceding claims, wherein the estimating device ( 10 ) based on the measured values ( 5 ) of several measuring cycles a respective time series ( 18 . 19 ) is formed for the at least one operating _variable (U _BATT , I _BATT ) and by means of a frequency analysis ( 20 ) of the at least one time series ( 18 . 19 ) a frequency-dependent impedance (Z) of the battery ( 2 ) is determined. Method according to one of the preceding claims, wherein the estimation device ( 10 ) is selected by means of a support vector machine (SVM) by a classification of a plurality of predetermined classes (c ₁ , c ₂ , c _n ), wherein each of the classes (c ₁ , c ₂ , c _n ) represents a state of charge. Method according to one of the preceding claims, wherein the state of charge value is only replaced by the estimated value, if in at least one previous measurement cycle the determined state of charge ( 8th ) in a predetermined state of charge ( 13 ) lies. Method according to one of the preceding claims, wherein the state of charge value is replaced by the estimated value only if the previous or the current covariance violates a predetermined quality criterion. Method according to one of the preceding claims, wherein the state of charge ( 8th ) of a lithium-ion battery is determined. Contraption ( 4 ) for a motor vehicle ( 1 ), with a measuring input for receiving respective measured values ( 5 ) at least one operating _variable (U _BATT , I _BATT , T) of a battery ( 2 ) of the motor vehicle ( 1 ) and a computing device comprising a Kalman filter device ( 9 ) for estimating a state of charge of the battery ( 2 ) and one of the Kalman filter device ( 9 ) independent estimator ( 10 ) for generating an estimate of the state of charge, wherein the device ( 4 ) is adapted to perform a method according to any one of the preceding claims. Motor vehicle ( 1 ) comprising: - a battery ( 2 ), - a measuring device ( 3 ) for measuring measured values ( 5 ) at least one operating _variable (U _BATT , I _BATT , T) of the battery ( 2 ) and - one with the measuring device ( 3 ) coupled device ( 4 ) according to claim 9.

Images