Method and Apparatus for Operating a System for Providing Predicted Confidence Intervals for States of Health of Electrical Energy Stores for at Least One Device With the Aid of Machine Learning Methods

Patent metadata

Field	Value
Publication number	US-2023016228-A1
Application number	US-202217812210-A
Country	US
Kind code	A1
Filing date	Jul 13, 2022
Priority date	Jul 14, 2021
Publication date	Jan 19, 2023
Grant date	—

Abstract

A computer-implemented method predicts a modeled state of health of an electrical energy store having at least one electrochemical unit in a technical device. The method includes providing a data-based state of health model, based on a characteristic of at least one operating variable of the electrical energy store up to a time, to assign the electrical energy store a corresponding state of health for the time and to indicate a corresponding modeling uncertainty, and predicting the characteristic of the at least one operating variable starting from a present time into the future based on a usage pattern model that is determined by a user-specific or usage-specific usage pattern. The method further includes predicting a characteristic of the state of health based on the data-based state of health model and the predicted characteristic, generated in a model-based manner, of the at least one operating variable.

First claim

What is claimed is:

1 . A computer-implemented method for predicting a modeled state of health of an electrical energy store having at least one electrochemical unit in a technical device, the method comprising:
providing a data-based state of health model, based on a characteristic of at least one operating variable of the electrical energy store up to a time, to assign the electrical energy store a corresponding state of health for the time and to indicate a corresponding modeling uncertainty; predicting the characteristic of the at least one operating variable starting from a present time into the future based on a usage pattern model that is determined by a user-specific or usage-specific usage pattern; predicting a characteristic of the state of health based on the data-based state of health model and the predicted characteristic, generated in a model-based manner, of the at least one operating variable; determining a confidence interval for the predicted characteristic of the state of health based on the modeling uncertainty of the data-based state of health model and a confidence of the usage pattern model; signaling the predicted characteristic of the state of health and the corresponding confidence interval.

2 . The method according to claim 1 , wherein, in order to determine the predicted characteristic of the state of health, the data-based state of health model is operated with an overall characteristic of the at least one operating variable that comprises a previous characteristic of the at least one operating variable up to the present time and the characteristic, generated in a model-based manner, of the at least one operating variable starting from the present time into the future.

3 . The method according to claim 1 , wherein the usage pattern model is configured as a Bayesian long short-term memory in order, based on usage parameters of the usage-specific usage pattern, to continuously provide the characteristic of the at least one operating variable or a characteristic of at least one loading variable from which the at least one operating variable is generated.

4 . The method according to claim 1 , wherein the confidence of the usage pattern model is determined based on an uncertainty of the prediction of the characteristic, generated in a model-based manner, for the at least one operating variable.

5 . The method according to claim 1 , wherein:
a confidence of the prediction of the at least one operating variable is ascertained based on characteristics of the at least one operating variable that are acquired in the past, the confidence of the usage pattern model is ascertained by comparing a time period of an actual characteristic of the at least one operating variable with the one or more predicted characteristics of the at least one operating variable for the corresponding time period, and the one or more predicted characteristics of the at least one operating variable is or are performed in each case for a constant forecast horizon.

6 . The method according to claim 5 , wherein:
the one or more predicted characteristics of the at least one operating variable are each formed by values of the at least one operating variable at forecast target times that are ascertained based on a respective usage pattern model that is trained with the actual characteristic of the at least one operating variable up to a corresponding forecast start time, and the forecast start times lie before the corresponding forecast target times by the corresponding constant forecast horizon.

7 . The method according to claim 5 , wherein the confidence of the usage pattern model is determined based on a probability density function of deviations of the actual characteristic of the at least one operating variable and one of the one or more predicted characteristics of the at least one operating variable for the same period.

8 . The method according to claim 7 , wherein:
the confidence of the usage pattern model is determined in the form of a characteristic of the deviations based on the actual characteristic of the at least one operating variable and the plurality of predicted characteristics of the at least one operating variable for the same period for different constant forecast horizons, deviations between the constant forecast horizons are interpolated in order to obtain the characteristic of the deviation based on the forecast time of the value, to be predicted, of the state of health.

9 . The method according to claim 5 , wherein:
the confidence interval for the predicted characteristic of the state of health is indicated by confidence interval limits that result from limit state of health characteristics, the limit state of health characteristics result from state of health characteristics for a minimum and a maximum loading of the electrical energy store that are obtained by applying the characteristics of the maximum deviations to the predicted characteristic of the at least one operating variable and that are each those that are furthest away from a nominal state of health trajectory, and the modeling uncertainty of the data-based state of health model is applied to the state of health characteristics thus obtained for the minimum and the maximum loading of the electrical energy store in order to obtain the limit state of health characteristics.

10 . The method according to claim 1 , wherein:
the state of health model is a hybrid model and comprises a physical ageing model based on electrochemical model equations and configured to output a physical state of health and a trainable data-based correction model, and the correction model is trained to correct the physical state of health and to provide the corrected physical state of health as the modeled state of health and the corresponding model uncertainty.

11 . The method according to claim 1 , wherein:
the electrical energy store is operated based on the characteristic of the predicted modeled state of health and based on the confidence interval, a remaining service life of the electrical energy store is signaled based on the characteristic of the predicted modeled state of health and/or a range of a possible remaining service life is signaled based on the characteristic of the predicted modeled state of health and the confidence interval, and the number of remaining permitted fast-charging cycles is increased or reduced or current and derating limits for operation of the electrical energy store are optimized based on the range of possible remaining service life of the electrical energy store.

12 . The method according to claim 1 , wherein predictive maintenance intervals of the technical device are adapted based on the confidence of the usage pattern model.

13 . The method according to claim 1 , wherein the electrical energy store is used to operate a device, such as a motor vehicle, a pedelec, an aircraft, a drone, a machine tool, a consumer electronics device, a cell phone, an autonomous robot, and/or a domestic appliance.

14 . The method according to claim 1 , wherein a computer program product comprises commands that, when the computer program product is executed by at least one data processing apparatus, cause the at least one data processing apparatus to perform the method.

15 . The method according to claim 14 , wherein a non-transitory machine-readable storage medium comprises commands that, when executed by at least one data processing apparatus, cause the at least one data processing apparatus to perform the method.

16 . An apparatus predicts a modeled state of health of an electrical en

Assignees

• Bosch Gmbh Robert

Inventors

• Simonis Christian

Classifications

• G01R31/367

  Software therefor, e.g. for battery testing using modelling or look-up tables · CPC title

• G01R31/392 Primary

  Determining battery ageing or deterioration, e.g. state of health · CPC title

• G01R31/396

  Acquisition or processing of data for testing or for monitoring individual cells or groups of cells within a battery · CPC title

• G01R31/378

  specially adapted for the type of battery or accumulator · CPC title

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title