Method and system for predicting battery health with machine learning model

Patent metadata

Field	Value
Publication number	US-12036890-B2
Application number	US-202016952097-A
Country	US
Kind code	B2
Filing date	Nov 19, 2020
Priority date	Nov 19, 2020
Publication date	Jul 16, 2024
Grant date	Jul 16, 2024

Abstract

A method and a system for predicting battery health based on distance driven on a full battery load with machine learning model are provided. The method includes: obtaining historical vehicle telematics of vehicles, wherein the historical vehicle telematics comprise at least one of the following: odometer readings, battery SOC, vehicle speed, battery-module temperatures, and battery-cell voltages; creating a distance driven model according to a relationship between a distance driven on a full battery load of vehicles and the historical vehicle telematics; obtaining a distance driven on a full battery load of a vehicle based on the distance driven model by using real-time vehicle telematics of the vehicle as model input; predicting battery health of the vehicle by comparing the obtained distance with a reference distance value.

First claim

What is claimed is:

1. A method for predicting battery health with machine learning model, comprising:
obtaining historical vehicle telematics of vehicles, wherein the historical vehicle telematics comprise at least one of the following: odometer readings, battery state-of-charge (SOC), vehicle speed, battery-module temperatures, and battery-cell voltages;
creating a distance driven model according to a relationship between a distance driven on a full battery load of vehicles and the historical vehicle telematics;
obtaining a distance driven on a full battery load of a vehicle based on the distance driven model by using real-time vehicle telematics of the vehicle as model input; and
predicting battery health of the vehicle by comparing the obtained distance with a reference distance value;
creating a distance driven model according to a relationship between a distance driven on a full battery load of vehicles and the historical vehicle telematics comprises:
identifying trips on a battery from the historical vehicle telematics of vehicles;
extracting at least one of the following model features from the historical vehicle telematics: positive/negative acceleration counts, a mean vehicle speed, a mean battery temperature, battery temperature imbalance, regenerated energy, a mean cell-voltage difference, and a cumulative distance; and
modeling the distance driven according to the relationship between the distance driven on a full battery load and the model features extracted from the historical vehicle telematics.

2. The method as claimed in claim 1 , after creating the distance driven model, the method further comprises:
training the distance driven model based on historical data of vehicles to provide a predictive general model; and
training individual-vehicle models for different vehicles by using predictive general model coefficients as a starting point.

3. The method as claimed in claim 1 , before modeling distance driven on a full battery load of vehicles, the method further comprises:
scaling the distance driven, the positive/negative acceleration counts and the regenerated energy for each of the trips based on SOC change.

4. The method as claimed in claim 1 , before obtaining a distance driven on a full battery load of a vehicle based on the distance driven model by using real-time vehicle telematics of the vehicle as model input, the method further comprises:
collecting the real-time vehicle telematics from the vehicle at a preset frequency.

5. The method as claimed in claim 1 , wherein obtaining the distance driven on a full battery load of a vehicle based on the distance driven model by using real-time vehicle telematics of the vehicle as model input, comprises at least one of the following:
obtaining changes in distance driven due to trip specifics according to the following model features: positive/negative acceleration counts, a mean vehicle speed, and regenerated energy;
obtaining changes in distance driven due to seasonal temperature fluctuations as claimed in the following model features: the mean battery temperature, and the battery temperature imbalance; and
obtaining changes in distance driven due to long-term loss of battery health according to the following model features: the mean cell-voltage difference, and the cumulative distance.

6. The method as claimed in claim 5 , wherein a loss of distance driven with time is proportional to a cumulative usage and the mean cell-voltage difference.

7. The method as claimed in claim 1 , wherein the relationship between the distance driven on a full battery load and the historical vehicle telematics is determined by the following linear regression formula:
y=θ t X

where y is a distance driven on in a trip, X is at least one of the model features from the historical vehicle telematics, θ is a corresponding model coefficient, and θ t is the matrix transpose of θ.

8. The method as claimed in claim 1 , after predicting battery health of the vehicle by comparing the predictive distance with a reference distance value, the method further comprises:
sending an alarm when the battery health of the vehicle is below a preset threshold.

9. A non-volatile computer readable storage medium, in which a program is stored, the program is configured to be executed by a computer to perform the method as claimed in claim 1 .

10. A non-volatile computer readable storage medium, in which a program is stored, the program is configured to be executed by a computer to perform the method as claimed in claim 2 .

11. A non-volatile computer readable storage medium, in which a program is stored, the program is configured to be executed by a computer to perform the method as claimed in claim 3 .

12. A non-volatile computer readable storage medium, in which a program is stored, the program is configured to be executed by a computer to perform the method as claimed in claim 4 .

13. A non-volatile computer readable storage medium, in which a program is stored, the program is configured to be executed by a computer to perform the method as claimed in claim 5 .

14. A non-volatile computer readable storage medium, in which a program is stored, the program is configured to be executed by a computer to perform the method as claimed in claim 6 .

Assignees

• Guangzhou Automobile Group Co

Inventors

• Rangel Tonatiuh
• Zhang Yizhen
• Ho Dao Hoang
• Fu Ying
• Xu Yonggang

Classifications

• G01R31/367

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

• B60L3/12 Primary

  Recording operating variables {; Monitoring of operating variables} · CPC title

• G01R31/382

  Arrangements for monitoring battery or accumulator variables, e.g. SoC · CPC title

• B60L58/12

  responding to state of charge [SoC] · CPC title

• G01R31/392

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