Method and apparatus for providing an aging state for a device battery having correction of state observations on the basis of systematic state and environmental influences

1 . A method for ascertaining an aging state of a device battery of a battery-operated device, comprising:

detecting curves of operating variables of the battery-operated device;

providing at least one load factor at a determination time of detecting the curves of the operating variables;

providing a correction model that maps a correction variable based on the at least one load factor;

ascertaining the aging state of the device battery by evaluating the curves of the operating variables based on (i) an aging state model, an aging state observer, or an aging state measurement, and (ii) the correction variable;

creating the correction model by:

receiving curves of operating variables of corresponding device batteries of a plurality of device batteries and at least one associated load factor at a particular determination time for each of the corresponding device batteries;

evaluating the received curves of operating variables of the corresponding device batteries within an evaluation period to ascertain, for each of the device batteries of the plurality of device batteries, a modeled aging state, of a plurality of modeled aging states, of the corresponding device battery at the particular determination time;

ascertaining, for each of the device batteries of the plurality of device batteries, a corresponding smoothed aging state function curve calculated from all aging states of the plurality of device batteries within the evaluation period across a corresponding calendar age of the device batteries of the plurality of device batteries;

determining residuals of the aging states of all of the device batteries for each particular determination time as a difference of the modeled aging state and a function value of the corresponding smoothed aging state function at the corresponding calendar age associated with the modeled aging state;

for the plurality of device batteries, providing residual data which associates residuals with the at least one load factor associated with a correspondingly underlying modeled aging state of the plurality of modeled aging states; and

creating the correction model based on the residual data, the correction model configured to map the correction variable dependent on or corresponding to the residuals, based on the at least one load factor; and

operating the battery-operated device using electrical energy from the device battery according to the ascertained aging state of the device battery.

2 . The method according to claim 1 , wherein:

the at least one load factor comprises one or more battery states and/or one or more environmental conditions to which the device battery is exposed during the ascertainment of the aging state, and

the at least one load factor comprises one or more of:

an ambient temperature for the time of determination of the corresponding aging state;

a battery temperature for at least one time or an average battery temperature during the determination of the corresponding aging state;

a starting state of charge and/or a final state of charge when determining the aging state during a charging process;

a specification of a non-linearity of a current curve or state of charge curve during the determination of the corresponding aging state or a partial specification of a relationship of a state-of-charge shift to an ampere-hour throughput (dSoC/dAh) during a charging process;

one or more no-load terminal voltages in a relaxed state after a sufficiently long rest phase at one or more predetermined temperatures;

a battery temperature at a predetermined no-load terminal voltage in the relaxed state after a sufficiently long rest phase;

the aging state of the device battery; and

one or more electrochemical state parameters of battery cells of the device battery.

3 . The method according to claim 1 , wherein the operating variables comprise at least one of battery current, battery voltage, battery temperature, and state of charge.

4 . The method according to claim 1 , wherein:

the aging state is provided as a capacity-related aging state or as a resistance-change-related aging state; and

the aging state model is configured to ascertain the capacity-related aging state by evaluating a battery current during a charging phase of the device batteries, using a coulomb counting method, or to ascertain the resistance-change-related aging state by evaluating a voltage gradient of a load jump which is greater than a predefined threshold value.

5 . The method according to claim 1 , wherein the corresponding smoothed aging state function corresponds to a linear function with a slope of 0 or less than 0, or corresponds to a curve of the aging states across the evaluation period, which curve is smoothed using a low-pass filter, provided as a PT1 or PT2 filter, or corresponds to smoothing by a sliding average value.

6 . The method according to claim 1 , wherein the correction model corresponds to a parametric model fitted with the residual data or to a data-based probabilistic regression model trained with the residual data, provided as a Gaussian process model or a Bayesian neural network.

7 . The method according to claim 6 , wherein:

when the probabilistic regression model is used as a correction model, a confidence for the ascertained correction variable is determined; and

the aging state is ascertained depending on the correction variable only when the confidence exceeds a predefined confidence threshold value.

8 . The method according to claim 1 , wherein:

providing the residual data and creating the correction model is carried out in a central unit remote from the device;

model parameters of the correction model are transmitted back to the device; and

the aging state of the device battery is determined in the battery-operated device or in the central unit.

9 . The method according to claim 8 , wherein a predicted aging state trajectory configured as a smoothed curve is determined based on the ascertained aging states to determine a remaining service life when a predefined aging state threshold is reached.

10 . The method according to claim 1 , wherein the correction model is created in a central unit remote from the device.

11 . The method according to claim 10 , wherein the dependence of the at least one load factor on the correction variable is evaluated and validated by a correlation coefficient or by a hypothesis test of a normal distribution in order to evaluate a quality of the correction model.

12 . An apparatus configured to carry out the method of claim 1 .

13 . A non-transitory machine-readable storage medium, comprising commands which, when executed by at least one data processing device, cause the data processing device to execute steps of the method according to claim 1 .