Systems and methods for predicting battery life using data from a diagnostic cycle

Patent metadata

Field	Value
Publication number	US-11768249-B2
Application number	US-202117218829-A
Country	US
Kind code	B2
Filing date	Mar 31, 2021
Priority date	Oct 29, 2020
Publication date	Sep 26, 2023
Grant date	Sep 26, 2023

Abstract

System, methods, and other embodiments described herein relate to improving the estimation of battery life. In one embodiment, a method includes measuring electrochemical data of a battery cell associated with an electrochemical reaction triggered by a test during a diagnostic cycle. The method also includes determining a feature associated with the degradation of the battery cell from the electrochemical data. The method also includes predicting an end-of-life (EOL) of the battery cell by using the feature in a machine learning (ML) model.

First claim

What is claimed is:

1. A prediction system comprising:
a memory communicably coupled to a processor and storing:
a prediction module including instructions that when executed by the processor cause the processor to:
measure data of a battery cell associated with an electrochemical reaction from a degradation mode having non-linear properties, and the degradation mode is triggered by tests during a diagnostic cycle;
identify a feature that correlates with degradation rates of the battery cell from the data by observations from varying discharge profiles and through grouping the tests according to physical properties, end-of-life (EOL) parameters, a series of chemical properties associated with the observations, and a number of cycles for the battery cell;
predict an EOL of the battery cell by using the feature in a machine learning (ML) model; and
cycle the battery cell according to the EOL for reducing the degradation rates.

2. The prediction system of claim 1 , wherein the prediction module includes instructions to predict the EOL further including instructions to process the feature by the ML model to discover an unknown quality of the battery cell that satisfies criteria for accuracy of the EOL.

3. The prediction system of claim 1 , wherein the prediction module includes instructions to identify the feature further including instructions to acquire information from measurements for the battery cell at operating conditions that vary to forecast the degradation rates.

4. The prediction system of claim 3 , wherein the operating conditions represent an envelope of the battery cell.

5. The prediction system of claim 3 , wherein the chemical properties are any one of a loss of active material (LAM), a loss of lithium inventory (LLI), and a retardation of internal kinetics associated with the battery cell.

6. The prediction system of claim 3 , wherein the prediction module further includes instructions to identify characteristics of the degradation rates measured from a point or spectrum of the information.

7. The prediction system of claim 3 , wherein the prediction module further includes instructions to correlate the feature to an energy fade or physical state of the battery cell and train the ML model using the feature.

8. The prediction system of claim 1 , wherein the feature is related to any one of a rate performance test (RPT), hybrid pulse power characterization (HPPC) resistance, and HPPC relaxation of the battery cell.

9. The prediction system of claim 1 , wherein the prediction module further includes instructions to determine a degradation condition internal to the battery cell according to the electrochemical reaction.

10. The prediction system of claim 1 , wherein the EOL is associated with a life span of a battery pack that includes the battery cell.

11. A non-transitory computer-readable medium comprising:
instructions that when executed by a processor cause the processor to:
measure data of a battery cell associated with an electrochemical reaction from a degradation mode having non-linear properties, and the degradation mode is triggered by tests during a diagnostic cycle;
identify a feature that correlates with degradation rates of the battery cell from the data by observations from varying discharge profiles through grouping the tests according to physical properties, end-of-life (EOL) parameters, a series of chemical properties associated with the observations, and a number of cycles for the battery cell;
predict an EOL of the battery cell by using the feature in a machine learning (ML) model; and
cycle the battery cell according to the EOL for reducing the degradation rates.

12. A method, comprising:
measuring data of a battery cell associated with an electrochemical reaction from a degradation mode having non-linear properties, and the degradation mode is triggered by tests during a diagnostic cycle;
identifying a feature that correlates with degradation rates of the battery cell from the data by observations from varying discharge profiles and through grouping the tests according to physical properties, end-of-life (EOL) parameters, a series of chemical properties associated with the observations, and a number of cycles for the battery cell;
predicting an EOL of the battery cell by using the feature in a machine learning (ML) model; and
cycling the battery cell according to the EOL for reducing the degradation rates.

13. The method of claim 12 , wherein predicting the EOL further comprises processing the feature by the ML model to discover an unknown quality of the battery cell that satisfies criteria for accuracy of the EOL.

14. The method of claim 12 , wherein
identifying the feature further includes acquiring information from measurements for the battery cell at operating conditions that vary to forecast the degradation rates.

15. The method of claim 14 , wherein the operating conditions represent an envelope of the battery cell.

16. The method of claim 14 , wherein the chemical properties are any one of a loss of active material (LAM), a loss of lithium inventory (LLI), and a retardation of internal kinetics associated with the battery cell.

17. The method of claim 14 , further comprising:
identifying characteristics of the degradation rates measured from a point or spectrum of the information.

18. The method of claim 14 , further comprising:
correlating the feature to an energy fade or physical state of the battery cell; and
training the ML model using the feature.

19. The method of claim 12 , wherein the feature is related to any one of a rate performance test (RPT), hybrid pulse power characterization (HPPC) resistance, and HPPC relaxation of the battery cell.

20. The method of claim 12 , further comprising:
determining a degradation condition internal to the battery cell according to the electrochemical reaction.

Assignees

• Toyota Res Inst Inc
• Univ Leland Stanford Junior
• Massachusetts Inst Technology

Inventors

• Chueh William C
• Van Vlijmen Bruis
• Gent William E
• Lam Vivek
• Herring Patrick K
• Balaji Gopal Chirranjeevi
• Asinger Patrick A
• Jiang Benben
• Braatz Richard Dean
• Cui Xiao
• Crane Gabriel B

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0985

  Hyperparameter optimisation; Meta-learning; Learning-to-learn · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G01R31/3842 Primary

  combining voltage and current measurements · CPC title