DE102021202904A1 - Method for monitoring an energy store in a motor vehicle - Google Patents

Abstract

The invention relates to a method for monitoring an energy store in a motor vehicle, with the energy store (10) supplying at least one consumer, preferably for a safety-relevant driving function, with at least one age-dependent variable (SOH(tpred)) of the energy store (10) for a future point in time ( tpred) is predicted, a predicted operating variable (Upred) of the energy store (10) being predicted at least as a function of the predicted age-dependent variable (SOH(tpred)) given a specific underlying load (42) of the energy store (10), the predicted operating variable (Upred) of the energy store (10) with a predeterminable limit value, in particular a minimum requirement such as reaching a minimum voltage (UBattmin), is compared, with the predicted age-dependent variable (SOH(tpred)) being based on at least one current and/or past age-dependent variable (SOH(t); SOH(t-1)) of energy p eichers (10), in particular derived from a measurement of a measured variable (U, I, T) such as voltage, current, temperature, etc. of the energy store (10) is determined.

Description

The invention relates to a method for monitoring an energy store in a motor vehicle according to the generic type of the independent claim.

State of the art

From the DE 102018220494 A1 a method for monitoring an energy store in a vehicle electrical system of a motor vehicle is already known, in which at least one current operating variable of the energy store is determined and this at least one operating variable is passed on to a forecast model and this forecast model uses the current value for the at least one operating variable to determine future values , wherein the future value is given to a voltage predictor, which calculates an expected minimum voltage of the energy store for a selected function. Battery condition detection software determines or measures the current capacity and internal resistance of the battery. These are passed on to a forecast model. The prognosis model calculates the future values of the capacity and the internal resistance with the help of representative load collectives and a load-carrying capacity model, on which the calculation of the expected minimum voltage is based.

The object of the invention is to further simplify the prediction of a specific parameter of the electrical energy store. This object is solved by the features of the independent claim.

Disclosure of Invention

Due to the fact that only at least one predicted age-dependent variable of the energy store is used to determine a predicted parameter of the energy store, which is only based on at least one current and/or past age-dependent variable, in particular without resorting to a complex prognosis model that is based, for example, on a load Based on a resilience model, the determination of the predicted parameter can be simplified. The computing effort and the required computing power can thus be reduced. In addition, the accuracy and robustness of the prediction increases. Furthermore, the failure rate of the energy store, which is particularly relevant for safety-critical applications, can be further reduced and its availability increased by means of predictive maintenance or predictive health management.

In an expedient development, it is provided that a remaining service life of the energy store is reached at the point in time when the predicted operating variable reaches the limit value, in particular the minimum requirement. This allows critical operating situations to be identified at an early stage and countermeasures such as predictive maintenance or the like to be initiated. Due to the more accurate forecast of the remaining service life, maintenance intervals can be planned more precisely or extended using predictive energy management, thus reducing maintenance costs, which is particularly advantageous for vehicle fleet operators. The safety of the vehicle can be increased further as a result.

In an expedient development, it is provided that, in order to determine the predicted age-dependent variable, a weighting, which in particular decreases exponentially over the past time, is carried out on at least the current and/or at least the past age-dependent variable. The memory requirements can be reduced, especially with an exponential weighting, since only the last measured value has to be saved for the prediction. The last age-dependent variable in each case is particularly preferably used or stored for the prediction of the predicted age-dependent variable.

In an expedient development, it is provided that at least one smoothing factor of the predicted age-dependent variable depends on a prediction error of the predicted operating variable and/or the predicted age-dependent variable. Especially when using a function-based end-of-life criterion, a further improvement in the accuracy of the forecast is possible through a corresponding adjustment of the smoothing parameters as a function of a one-step prediction error. As a result, the error between the extrapolated and quasi-measured course, in particular of the operating variable, can be reduced and the accuracy of the forecast of the remaining service life can thus be further increased.

A further expedient development provides that at least one smoothing factor of the predicted age-dependent variable is adapted as a function of the remaining service life, in particular by smoothing being reduced by the smoothing factor as the remaining service life increases. As a result, when the estimated remaining service life is high, the global or long-term trend and, when the end of life is approaching, the local or short-term trend of the time profiles of the aging-dependent variables is used for the prediction and thus the average accuracy of the prognosis of the Increased the remaining lifetime over the prediction horizon and reduced its variance.

In an expedient development, it is provided that the predicted operating variable of the energy store is determined using a model of the energy store and/or using a load profile. This allows the functionality to be determined individually for each function with particular precision.

In an expedient development, it is provided that at least one level and/or a trend of the predicted age-dependent variable is determined for the prediction of the predicted age-dependent variable. It is particularly preferably provided that the level is determined as a function of a current and/or past aging-dependent variable and/or as a function of a smoothing factor and/or of a previous level and/or as a function of a previous trend and/or that the trend is determined as a function of a current and/or previous level and/or a further smoothing factor and/or a previous trend. A robust and accurate prediction can thus be achieved in a particularly simple manner.

In an expedient development, it is provided that at least one of the smoothing factors is determined using an exponential smoothing method. This increases the accuracy and robustness of the method. A gentle adaptation of the smoothing factors can be achieved.

In an expedient development, it is provided that at least one of the smoothing factors is influenced by such an aging-dependent variable, which has a strong effect on the predicted operating variable at the current operating point of the energy store, being used for greater adaptation, in particular expressed by a gradient of the predicted operating variable regarding the age-dependent size. In this way, the error between the predicted and measured course of the operating variables can be reduced by a targeted, stronger adjustment of the corresponding course of the operating variables, and the prognosis of the remaining service life can thus become more accurate.

In an expedient development it is provided that at least one of the smoothing factors is influenced as a function of the remaining service life of the energy store, in particular by an exponential dependency of the smoothing factor on an estimated value of the remaining service life. This further increases the accuracy and robustness.

In a particularly expedient manner, the comparison of the predicted operating variable with the limit value is carried out for certain other parameters such as the specified state of charge and/or temperature. As a result, certain minimum requirements are met under defined, comparable conditions.

A device is particularly preferably set up to carry out the method, in particular at least part of a sensor, preferably a battery sensor, and/or a control unit of a vehicle, in particular a vehicle computer, gateway, energy management control unit and/or a computer of a cloud system. Depending on the application and the computing power available, the method can be implemented in the appropriate device.

Further expedient developments result from further dependent claims and from the description.

character list

• 1 eine Implementierung der Vorhersage einer Kenngröße des Energiespeichers in einem Sensor, insbesondere Batteriesensor,
• 2 eine Implementierung der Vorhersage einer Kenngröße des Energiespeichers in einem Fahrzeugcomputer,
• 3 eine Implementierung der Vorhersage einer Kenngröße des Energiespeichers in einer Cloud,
• 4 ein Ablaufdiagramm des erfindungsgemäßen Verfahrens,
• 5 einen zeitabhängigen SOH-Verlauf für eine geglättete Kurve im Vergleich zur Messung sowie
• 6 einen zeitabhängigen SOF-Verlauf, berechnet aus dem geglätteten SOH-Verlauf im Vergleich zur Messung.

Show it

• 1 an implementation of the prediction of a parameter of the energy store in a sensor, in particular a battery sensor,
• 2 an implementation of the prediction of a parameter of the energy storage in a vehicle computer,
• 3 an implementation of the prediction of a parameter of the energy storage in a cloud,
• 4 a flowchart of the method according to the invention,
• 5 a time-dependent SOH course for a smoothed curve compared to the measurement as well as
• 6 a time-dependent SOF curve, calculated from the smoothed SOH curve compared to the measurement.

Embodiments of the invention

The invention is shown schematically in the drawings using embodiments and is described in detail below with reference to the drawings.

A battery or accumulator is described as an example of a possible energy store in the exemplary embodiment. Alternatively, however, other energy storage devices suitable for this task can be based, for example, on inductive or kapa zitive base, fuel cells, capacitors or the like are equally used.

In the embodiment according to 1 an energy store 10 is connected to a sensor 12 . The sensor 12 detects certain measured variables of the energy store 10 such as current I, voltage U, temperature T. In addition, a state detection 13 is implemented in the sensor 12 . At least one state variable of the energy store 10 can be determined via the state detection 13 with recourse to the recorded measured values I, U, T. For example, the current age-dependent variable of the energy store 10 as SOH(t) (State of Health) and/or a state of charge SOC (State of Charge) and/or a functional state SOF (State of Function) and/or a remaining service life RUL (Ramaining Useful Life). For example, the internal resistance Ri, the capacitance C0, transmission polarization or the like can be used as age-dependent variables SOH of the energy store 10 . The age-dependent variables SOH are characterized in that they change depending on the age of the energy store 10 . The aging-dependent variables SOH can preferably be detected by the sensor 12 via the measured variables UIT and can be calculated via the status detector 13 . The state recognition 13 preferably works continuously. The remaining service life RUL made available by the sensor 12 is then possibly transmitted together with other prognosis or diagnostic variables via a communication interface (a bus system such as LIN, CAN) to higher-level systems (not shown) for further processing, for example for the purpose of predictive maintenance and/or or a predictive or preventive health management of the energy store 10.

The embodiment according to 2 differs from the one after 1 now in that the sensor 12 or the status detection 13 only generates certain parameters such as SOC, SOF, SOH, but not the remaining service life RUL. The remaining service life RUL is determined outside of sensor 12, for example in other (vehicle) components 24 with sufficient computing resources, such as a vehicle computer, a gateway, a control unit or another computing unit. For this purpose, the characteristic values of the energy store 10 such as SOC, SOF, SOH determined in the sensor 12 are transmitted to the further component 24 . The transmission takes place, for example, via appropriate communication systems, for example a LIN interface.

In 3 the method can also be implemented outside of the vehicle in a cloud-based system if specific parameters such as the SOH data of the energy store 10 are continuously sent from the vehicle 31, 32, 33 to a cloud 26 via a preferably wireless data connection 28. The corresponding algorithm for determining the remaining service life RUL is stored in the cloud system 26 . There, for different vehicles 31, 32, 33, different remaining lifetimes RUL1, RUL2, RUL3 of the respective energy stores 10.1, 10.2, 10.3 can be determined. This solution is particularly interesting for fleet operators, for example for remote maintenance. In addition, with this arrangement, the prediction quality can be further increased with the aid of the information transmitted from other fleet vehicles to the cloud 26, such as vehicle use or climatic conditions. Furthermore, future adjustments to the method for determining the remaining service life, including an end-of-life criterion (EOL, End of Life), can be implemented very easily.

In 4 a flow chart of the method for forecasting the remaining service life RUL of the energy store 10 is shown in more detail.

To predict the expected remaining service life RUL, the method described below uses a specific extrapolation method for the linear extrapolation of size or level L and trend T of the measured time curves of the age-dependent sizes SOH of the energy store 10, such as capacitance Co and/or internal resistance Ri, until a predeterminable or predetermined end-of-life criterion (EOL) in order to determine the remaining service life RUL. This end-of-life criterion can be mapped, for example, via a limit value UBattmin if, for example, certain operating variables U or predicted operating variables Upred of the energy store 10 no longer reach certain minimum values. The method particularly preferably uses what is known as “Double Exponential Smoothing” or also known as “Holt's Linear Trend Method” for the linear extrapolation of the measured time curves. In this case, the smoothing factors a, β are continuously adjusted as a function of the remaining service life RUL or an estimated value of the remaining service life RUL. The adjustment is made in such a way that with a high estimated remaining service life RUL, strongly smoothed time curves of the age-dependent variables SOH are used for the prediction. The smoothing is continuously reduced as the end of life EOL is approached. As a result, when the estimated remaining service life RUL is high, the global or long-term trend T and/or when the estimated remaining service life RUL approaches the end of life EOL, the local or short-term trend T of the time profiles of the aging-dependent variables SOH for predicting the predicted age tion-dependent variable SOHpred is used. This increases the average accuracy of the prediction of the remaining lifetime RUL over the prediction horizon, while the variance is reduced.

When using an end-of-life criterion EOL, which is based on the functional state SOF of the energy store 10, a further improvement in the prediction accuracy can be achieved by adapting the smoothing factors a, β depending on a so-called one-step prediction error, in particular the SOF prediction. As a result, the error between the extrapolated and quasi-measured SOF curve can be reduced and the accuracy of the forecast of the remaining service life RUL can thus be further increased.

The input variables of the method are measured or estimated time curves that describe the energy store 10, such as the current state of health (=SOH(t)) or age-dependent variable or variables that describe it, such as capacitance C0, internal resistance Ri, polarizations, etc. of the considered energy store 10, the z. B. from the state detection 13 of the sensor 12, in particular battery sensor, are continuously determined.

Using the time series prediction, age-dependent variables for a future point in time tpred=t+RUL are calculated on the basis of past and current input variables SOH(t-1), SOH(t) (SOH(tpred)), block 40.

Then, by means of a predictor 44, in particular a SOF predictor, based on a mathematical model 49 of the energy store 10, the predicted operating variable Upred of the energy store 10, such as the voltage drop to be expected with regard to a predefined application-specific load profile, such as a current load profile ILoad(t) ( Block 42) is calculated and thus checked whether the predicted aging state or predicted aging-dependent variable SOH(tpred) and specified state of charge SOC0 and temperature TBatt0 fall below a limit value required for the application, such as a minimum voltage UBattmin. Here, for example, an equivalent circuit diagram or model 49 of the energy store 10 can be used using certain operating variables such as current I, voltage U, temperature T and state variables such as the open-circuit voltage, the concentration polarization, the passage polarization of the various electrodes etc. as well as corresponding parameters (e.g Internal resistance Ri, equivalent capacity C0, resistance and capacity of the acid diffusion or resistance and capacity of the double layer of the electrodes, etc.

und damit die Restlebensdauer RUL bestimmt, Block 48.The solution tpred of the non-linear equation is then iteratively determined using a zero-point search method such as Newton's or secant's method $$\begin{array}{l}\text{updated}\left(\text{ILoad},\text{SOH}\left(\text{tpred}\right),\text{<figure-callout id="0" label="SOC" filenames="DE102021202904A1\_0025.png,DE102021202904A1\_0026.png" state="{{state}}">SOC</figure-callout>}\mathrm{0,}\text{Tbatt}0\right)-\text{UBatt},\text{at least}=0\\ =>\text{tpred}=>\text{RUL}=\text{tpred}-\text{t}\end{array}$$

and thus the remaining service life RUL is determined, block 48.

The time series prediction according to block 40 is carried out as described below. A special smoothing method, in particular the so-called double exponential smoothing method, is used to predict a predicted age-dependent variable SOH(tpred) on the basis of current and past age-dependent variables SOH(t-1), SOH(t), the level L and trend T of the measured SOH(t) curves are determined using the following equations and thus future predicted age-dependent variables SOH(tpred) are linearly extrapolated:

Trend (T) Update (des SOH-Werts): $${\text{T}}_{\text{t}}=\mathrm{\beta }\left[{\text{L}}_{\text{t}}-{\text{L}}_{\text{t}-1}\right]+\left(1-\mathrm{\beta }\right){\text{T}}_{\text{t}-1}$$

• Y_t: SOH Messwert zum Zeitpunkt t
• Ŷ_t : SOH Vorhersagewert zum Zeitpunkt t
• α : Glättungsfaktor für den Level L
• β : Glättungsfaktor für den Trend T 1-Schritt Vorhersagewert:

$${\widehat{\text{Y}}}_{\text{t}}={\text{L}}_{\text{t}-1}+{\text{T}}_{\text{t}-1}$$

m-Schritt Vorhersagewert: $${\widehat{\text{Y}}}_{\text{t}+\text{m}}={\text{L}}_{\text{t}}+{\text{mT}}_{\text{t}}$$

Level (L) update (of the SOH value or the age-dependent variable): $${\text{L}}_{\text{t}}=\mathrm{a}{\text{Y}}_{\text{t}}+\left(1-\mathrm{a}\right)\left[{\text{L}}_{\text{t}-1}+{\text{T}}_{\text{t}-1}\right]$$

Trend (T) update (of the SOH value): $${\text{T}}_{\text{t}}=\mathrm{\beta }\left[{\text{L}}_{\text{t}}-{\text{L}}_{\text{t}-1}\right]+\left(1-\mathrm{\beta }\right){\text{T}}_{\text{t}-1}$$

• Y _t : SOH measured value at time t
• Ŷ _t : SOH predicted value at time t
• α : smoothing factor for level L
• β : smoothing factor for the trend T 1-step predicted value:

$${\widehat{\text{Y}}}_{\text{t}}={\text{L}}_{\text{t}-1}+{\text{T}}_{\text{t}-1}$$

m-step predicted value: $${\widehat{\text{Y}}}_{\text{t}+\text{m}}={\text{L}}_{\text{t}}+{\text{mT}}_{\text{t}}$$

The calculation of the smoothing factors α and β in block 40 is described below. The appropriate choice of the smoothing factors α and β is decisive for the accuracy and robustness of the method. Here, the Smooth Transition Exponential Smoothing (STES) approach according to James W. Taylor was chosen as an example, as it allows more stable predictions. STES enables smooth adaptation of the smoothing factors α, β depending on a transition variable Vt, which is usually specified as a function of the one-step prediction error.

$${\beta }_t=\frac{1}{1+\text{exp}\left(b_1+b_2{V}_t\right)}$$

$$V_t=e_t{}^2$$

V_t: Transitionsvariable $$e_t={\widehat{Y}}_t-Y_t$$

e_t: Einschrittvorhersagefehler $$a_1,a_2,b_1,b_2$$

Empirische Konstanten mit a₂,b₂ < 0Calculation of the smoothing factors according to STES: $$a_t=\frac{1}{1+\text{ex}\left(a_1+a_2{V}_t\right)}$$

$${\beta }_t=\frac{1}{1+\text{ex}\left(b_1+b_2{V}_t\right)}$$

$$V_t=e_t{}^2$$

V _t : transition variable $$e_t={\widehat{Y}}_t-Y_t$$

e _t : one-step prediction error $$a_1,a_2,b_1,b_2$$

Empirical constants with a ₂ ,b ₂ < 0

Improved RUL forecast accuracy on SOF EOL criterion

When using a SOF EOL criterion (function-specific load lload(t)), the determination is based on the predictive operating variable Upred and compared with a limit value, for example minimum voltage Ubattmin, as described in block 44, 46 as an example), the forecast accuracy can be increased by subsequent expansion of the transition variables are increased by an additional additive term: $$V_t=e_t{}^2+\text{fac}\frac{\text{dSOF}}{\text{dSOH}}\left(t\right)e_t{}^2$$

The additional term causes aging-dependent variables SOH, which have a strong effect on the predicted SOF value or the predicted operating variable Upred at the current operating point of the battery or the energy store 10, expressed by the SOF gradient with respect to SOH, to a greater extent Adaptation of the smoothing factors a, β and thus the error between the predicted and "measured" SOF curve or operating variable U is reduced through targeted stronger adaptation of the corresponding curve of the aging-dependent variable SOH and the RUL forecast becomes more accurate (cf. 5 ). In 5 the smoothed curve 54 of the age-dependent variable SOH is plotted in comparison to the measured curve 52 of the age-dependent variable SOH.

An improved RUL prognosis accuracy by adapting the smoothing factors α, β depending on the remaining service life is achieved as described below. If the remaining service life RUL is assumed to be large, a forecast of the remaining service life RUL on the basis of local trends T of the age-dependent variable SOH or local SOH trends, i.e. weak or unsmoothed curves of the age-dependent variable SOH(t), does not make sense, as this would result in too great a variance in the linear extrapolation determined values of the remaining service life RUL. On the other hand, the local trend L of the age-dependent quantity or the local SOH trend T near the end of life EOL is decisive for an exact determination of the remaining service life RUL. A remaining service life-dependent smoothing of the measured SOH(t) curves therefore leads to an improvement in the accuracy and robustness of the forecast of the remaining service life RUL over the prediction horizon as in 6 shown. Curve 58 designates the progression of the operating variable U or SOF calculated from a smoothed progression of the age-dependent variable SOH in comparison to the measurement, curve 56. The predicted EOL point in time tpred is also plotted. As described, there is less smoothing near the end of life SOL or tpred.

$${\beta }_t=\text{max}\left(\mathrm{1,}{d}_1\text{ exp}\left(d_{2}RU{L}_{est}\right)\right)\frac{1}{1+\text{exp}\left(b_1+b_2{V}_t\right)}$$

mit c₁, d₁ > 1, c₂, d₂ < 0, RUL_est = (U_Batt,pred(t) - U_Batt,min)/(U_Batt,pred(t) - U_Batt,pred(t=0))*tA method and an arrangement for predicting the remaining service life RUL of electrical energy stores 10 are described below. The adjustment of the smoothing factors α, β depends on the RUL estimated value RUL _est : $$a_t=\text{Max}\left(\mathrm{1,}{c}_1\text{ex}\left(c_{2}Ru{L}_{est}\right)\right)\frac{1}{1+\text{ex}\left(a_1+a_2{V}_t\right)}$$

$${\beta }_t=\text{Max}\left(\mathrm{1,}{\mathrm{i.e}}_1\text{ex}\left({\mathrm{i.e}}_{2}Ru{L}_{est}\right)\right)\frac{1}{1+\text{ex}\left(b_1+b_2{V}_t\right)}$$

with c ₁ , d ₁ > 1, c ₂ , d ₂ < 0, RUL _est = (U _Batt,pred (t) - U _Batt,min )/(U _Batt,pred (t) - U _Batt,pred (t =0))*t

The RUL estimated value RUL _est is linearly extrapolated from the predicted voltages U _pred (SOH(t ₁ )), U _pred (SOH(t ₂ )) calculated at two different times t ₁ , t ₂ using the SOF predictor 44 .
The prerequisite is that: U _pred (t ₂ ) - U _pred (t ₁ ) << 0 for t ₂ > t ₁ . Here t ₁ =0 and t ₂ =t was chosen.

By defining an exponential dependence of the smoothing factors a, β on the estimated value of the remaining service life RULest, the influence of the remaining service life RUL on the smoothing factors α, β is limited to a range close to EOL. The parameters of the functions for the smoothing factors α, β can be determined, for example, by optimization methods with the aim of minimal variance of the RUL forecast error from SOH(t) curves measured up to EOL.

If query 46 reveals that the remaining service life RUL has been reached, countermeasures can be initiated in block 50 . Corresponding warnings can be generated for the driver, vehicle operator, workshop or the like. Predictive maintenance could be initiated. Certain security-relevant functions can also be blocked or no longer permitted. It would also be possible to initiate appropriate countermeasures such as driving to the next parking lot, stopping at the hard shoulder, etc. in the event of particularly serious errors.

The method described is particularly suitable for monitoring energy storage devices 10 for safety-relevant applications such as, for example, for supplying safety-relevant consumers in a motor vehicle, in particular in the case of autonomous driving. However, the use is not limited to this.

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• DE 102018220494 A1 [0002]

Claims (15)

Method for monitoring an energy store in a motor vehicle, the energy store (10) supplying at least one consumer, preferably for a safety-related driving function, with at least one age-dependent variable (SOH(tpred)) of the energy store (10) being predicted for a future point in time (tpred). , a predicted operating variable (Upred) of the energy store (10) being predicted at least as a function of the predicted age-dependent variable (SOH(tpred)) given a specific underlying load (42) of the energy store (10), the predicted operating variable (Upred) of the Energy store (10) is compared with a predeterminable limit value, in particular a minimum requirement such as reaching a minimum voltage (UBattmin), characterized in that the predicted age-dependent variable (SOH(tpred)) is derived from at least one current and/or past age-dependent variable ( SOH(t); SOH(t-1)) of energy storage hers (10), in particular derived from a measurement of a measured variable (U, I, T) such as voltage, current, temperature, etc. of the energy store (10). procedure after claim 1 , characterized in that a remaining service life (RUL) of the energy store (10) is reached at the point in time when the predicted operating variable (Upred) reaches the limit value, in particular the minimum requirement (UBattmin). Method according to one of the preceding claims, characterized in that for determining the predicted age-dependent variable (SOH (tpred)), a weighting, which in particular decreases exponentially over the past time, is used at least for the current and/or at least the past age-dependent variable (SOH(t), SOH(t-1)) is made. Method according to one of the preceding claims, characterized in that at least one smoothing factor (α, β) of the predicted age-dependent variable (SOH (tpred)) depends on a prediction error of the predicted operating variable (Upred) and/or the predicted age-dependent variable (SOH (tpred) ) depends. Method according to one of the preceding claims, characterized in that at least one smoothing factor (α, β) of the predicted age-dependent variable (SOH(tpred)) is adapted as a function of the remaining service life (RUL), in particular by smoothing by the smoothing factor as the remaining service life increases (α, β) is reduced. Method according to one of the preceding claims, characterized in that the last age-dependent variable (SOH(t-1)) in each case is used or stored for the prediction of the predicted age-dependent variable (SOH(tpred)). Method according to one of the preceding claims, characterized in that the predicted operating variable (Upred) of the energy store (10) is determined using a model (49) of the energy store (10) and/or using a load profile (42). Method according to one of the preceding claims, characterized in that for the prediction of the predicted age-dependent variable (SOH(tpred)) at least one level (L) and/or a trend (T) of the predicted age-dependent variable (SOH(tpred)) is determined . Method according to one of the preceding claims, characterized in that the level (Lt) as a function of a current and/or past age-dependent variable (SOH(t), SOH(t-1)) and/or as a function of a smoothing factor (a ) and/or from a past level (Lt-1) and/or as a function of a past trend (Tt-1), in particular according to the following formula: Lt = α Yt + (1 - α) [Lt -1 + Tt -1], with Lt level, α smoothing factor, Yt age-dependent variable, Lt -1 past level, Tt-1 past trend. Method according to one of the preceding claims, characterized in that the trend (Tt) as a function of a current and/or previous level (Lt, Lt -1) and/or a further smoothing factor (β) and/or a previous trend (Tt -1) is determined, in particular according to the following formula: $$\text{tt}=\mathrm{\beta }\left[\text{Lt}-\text{Lt}-1\right]+\left(1-\mathrm{\beta }\right)\text{tt}-\mathrm{1,}$$

with Tt the current trend, β the additional smoothing factor, Lt the current level, Lt -1 the previous level, Tt -1 the previous trend. Method according to one of the preceding claims, characterized in that at least one of the smoothing factors (α, β) is determined using an exponential smoothing method, in particular using one of the following forms: $$a_t=\frac{1}{1+\text{ex}\left(a_1+a_2{V}_t\right)}$$

and or $${\beta }_t=\frac{1}{1+\text{ex}\left(b_1+b_2{V}_t\right)}$$

Where at is the smoothing factor at time t, βt is the further smoothing factor at time t, where vt is a transition variable or a square of a one-step prediction error (et) or the formula $$V_t=e_t{}^2$$

With $$e_t={\widehat{Y}}_t-Y_t$$

corresponds to where Yt represents the predicted age-dependent variable (SOH(tpred) and Yt represents the associated actual age-dependent variable (SOH(t)) and wherein a1, a2, b1, b2 represent empirical constants. Method according to one of the preceding claims, characterized in that at least one of the smoothing factors (α, β) is influenced by such an aging-dependent variable (SOH) which has a strong impact on the predicted operating variable (Upred) at the current operating point of the energy store (10). affects, is used for a stronger adjustment, expressed in particular by a gradient of the predicted operating variable (Upred) with respect to the aging-dependent variable (SOH), expressed in particular by the following formula: $$V_t=e_t{}^2+\text{fac}\frac{\text{dSOF}}{\text{dSOH}}\left(t\right)e_t{}^2$$

where vt is a transition variable and et is a one-step prediction error. Method according to one of the preceding claims, characterized in that at least one of the smoothing factors (α, β) is influenced as a function of the remaining service life (RUL) of the energy store (10), in particular by an exponential dependence of the smoothing factor (α, β) on a Estimated value of the remaining service life (RUL), in particular as a function of at least one of the following formulas $$a_t=\text{Max}\left(\mathrm{1,}{c}_1\text{ex}\left(c_{2}Ru{L}_{est}\right)\right)\frac{1}{1+\text{ex}\left(a_1+a_2{V}_t\right)}$$

$${\beta }_t=\text{Max}\left(\mathrm{1,}{\mathrm{i.e}}_1\text{ex}\left({\mathrm{i.e}}_{2}Ru{L}_{est}\right)\right)\frac{1}{1+\text{ex}\left(b_1+b_2{V}_t\right)}$$

with parameter c1, d1 > 1, parameter c2, d2 < 0, predicted remaining service life RULest = (UBatt,pred(t) - UBatt,min)/(UBatt,pred(t) - UBatt,pred(t=0)))* t. Method according to one of the preceding claims, characterized in that the comparison of the predicted operating variable (Upred) with the limit value is carried out with certain further parameters such as the specified state of charge (SOC0) and/or temperature (TBatt0). Device set up to carry out the method according to one of the preceding claims, the device in particular being at least part of a sensor (12), preferably a battery sensor, and/or a control unit of a vehicle, in particular a vehicle computer (24), gateway, energy management control unit and /or a computer of a cloud system (26).

Images