Method and apparatus for operating a system for providing predicted states of health of electrical energy stores for a device using machine learning methods

Patent metadata

Field	Value
Publication number	US-11965935-B2
Application number	US-202117537847-A
Country	US
Kind code	B2
Filing date	Nov 30, 2021
Priority date	Dec 3, 2020
Publication date	Apr 23, 2024
Grant date	Apr 23, 2024

Abstract

The invention relates to a computer-implemented method for predicting a modeled state of health of an electrical energy store having at least one electrochemical unit, in particular a battery cell, having the following steps: providing a data-based state of health model trained to assign a modeled state of health to the electrochemical energy store on the basis of characteristics of operating variables of the energy store; providing time characteristics of the operating variables that characterize operation of the electrical energy store; and determining a present or predicted modeled state of health on the basis of the generated characteristics of the operating variables using the state of health model, wherein data gaps in the time characteristics of the operating variables owing to a phase of inactivity are completed based on a characteristic of a temperature of the energy store that is derived from at least one provided ambient condition.

First claim

What is claimed is:

1. A method, which is computer-implemented, for predicting a modeled state of health of an electrical energy store having at least one electrochemical battery cell, the method comprising:
providing a data-based state of health model that is trained to assign a modeled state of health to the electrical energy store based on characteristics of operating variables of the electrical energy store, the data-based state of health model operating on a processor;
providing time characteristics of the operating variables that characterize operation of the electrical energy store to the provided data-based state of health model;
identifying a phase of inactivity of the electrical energy store in which no charging and no discharging of the electrical energy store has occurred, based on the time characteristics of the operating variables;
completing data gaps in the time characteristics of the operating variables during the identified phase of inactivity based on a characteristic of a temperature of the electrical energy store that is derived from at least one provided ambient condition to generate time characteristics of artificial operating variables;
determining a modeled state of health of the electrical energy store based on the provided time characteristics of the operating variables and the time characteristic of the artificial operating variables using the provided data-based state of health model operating on the processor, the determined modeled state of health being one of a present state of health and a predicted state of health; and
operating the electrical energy store to extend a remaining life of the electrical energy store based on the determined modeled state of health,
wherein the time characteristics of the operating variables are captured in a fast time frame between 0.1 Hz and 100 Hz,
wherein each of the time characteristics of the operating variables and the time characteristics of the artificial operating variables include a current, a voltage, a temperature, and a state of charge of the electrical energy store, and
wherein the data-based state of health model includes a purely data-based state of health model or a hybrid state of health model.

2. The method according to claim 1 further comprising:
ascertaining the characteristic of the temperature during the identified phase of inactivity using a data-based temperature model operating on the processor that is trained to provide a temperature of the electrical energy store based on at least one ambient condition.

3. The method according to claim 2 , wherein the data-based temperature model is configured to model the temperature of the electrical energy store based on (i) a location of the electrical energy store and (ii) a manner in which the electrical energy store is exposed to the at least one ambient condition.

4. The method according to claim 3 , wherein the location is a geographical location.

5. The method according to claim 2 , wherein the data-based temperature model is trained to assign the at least one ambient condition to the temperature of the electrical energy store.

6. The method according to claim 5 , wherein training data for training the data-based temperature model are generated by determining a steady-state ambient condition and a steady-state temperature of the electrical energy store at the time of an end of the identified phase of inactivity.

7. The method according to claim 2 , wherein the at least one ambient condition is one of (i) derived from weather information from a weather database and (ii) derived by measuring the at least one ambient condition directly.

8. The method according to claim 2 , wherein the at least one ambient condition is at least one of (i) an ambient temperature, (ii) an air pressure, and (iii) a humidity.

9. The method according to claim 1 further comprising:
modeling, based on a physical cooling model of the electrical energy store, the characteristic of the temperature during a predefined initial phase of the identified phase of inactivity,
wherein an initial battery temperature at a start of the identified phase of inactivity is provided to the physical cooling model, and
wherein the predefined initial phase ends when the modeled characteristic of the temperature is within a tolerance threshold of a predetermined steady state battery temperature.

10. The method according to claim 1 , wherein the temperature of the electrical energy store is a load variable for a dynamic model configured to model a voltage of the electrical energy store based on a current during the identified phase of inactivity and the temperature of the electrical energy store, the dynamic model being one of (i) in a form of an equivalent circuit model of the electrical energy store, (ii) in a form of an electrochemical model, and (iii) in a form of a single particle model.

11. The method according to claim 10 further comprising:
adapting the dynamic model based on the determined modeled state of health.

12. The method according to claim 11 , the adapting the dynamic model further comprising:
updating one of (i) model parameters and (ii) model states of the dynamic model based on the determined modeled state of health.

13. The method according to claim 10 , wherein:
the electrical energy store is a battery; and
the dynamic model provides a state of charge characteristic in the identified phase of inactivity based on the determined modeled state of health in accordance with a linear state of charge characteristic between a state of charge at an onset of the phase of inactivity and a state of charge at an end of the phase of inactivity.

14. The method according to claim 1 , wherein the electrical energy store is a battery.

15. The method according to claim 1 , wherein the data-based state of health model is in a form of the hybrid state of health model and includes (i) a physical health model that is based on electrochemical model equations and is configured to output a physical state of health of the electrical energy store, and (ii) a trainable data-based correction model that is trained to correct the physical state of health and to provide a corrected physical state of health as the determined modeled state of health.

16. The method according to claim 15 , wherein the trainable data-based correction model is at least one of (i) in a form of a regression model and (ii) in a form of a Gaussian process.

17. The method according to claim 1 , wherein the electrical energy store is used to operate one of (i) a motor vehicle, (ii) a pedelec, (iii) an aircraft, (iv) a drone, (v) a machine tool, (vi) a consumer electronic device, (vii) a cell phone, (viii) an autonomous robot, and (ix) a domestic appliance.

18. The method according to claim 1 , wherein a computer program has instructions that are executed by the processor to cause the processor to carry out the method.

19. An apparatus for predicting a modeled state of health of an electrical energy store having at least one electrochemical battery cell, comprising:
a processor configured to:
provide a data-based state of health model that is trained to assign a modeled state of health to the electrical energy store based on characteristics of operating variables of the electrical energy store;
provide time characteristics of the operating variables that characterize operation of the electrical energy store to the provided data-based state of health model;
identify a phase of inactivity of the electrical energy store in which no charging and no discharging of the electrica

Assignees

• Bosch Gmbh Robert

Inventors

• Simonis Christian
• Kroener Christoph

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G01R31/392 Primary

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

• G01R31/367 Primary

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

• G06N20/00

  Machine learning · CPC title