Self-adaptive lithium-ion battery method using knowledge-reinforced machine learning and kalman filtering, electronic device, and storage medium

Patent metadata

Field	Value
Publication number	US-12535783-B2
Application number	US-202318365621-A
Country	US
Kind code	B2
Filing date	Aug 4, 2023
Priority date	Aug 4, 2023
Publication date	Jan 27, 2026
Grant date	Jan 27, 2026

Abstract

The present disclosure provides a self-adaptive lithium-ion battery method using knowledge-reinforced machine learning and Kalman filtering, an electronic device, and a storage medium. The method includes training and synchronizing an artificial neural network with dual extended Kalman filters to capture battery capacity data of each of lithium-ion batteries; integrating prior knowledge with Gaussian process regression to form an integrated knowledge-reinforced Gaussian process regression; training a stochastic capacity degradation model by employing integrated knowledge-reinforced Gaussian process regression with captured battery capacity data to obtain a trained stochastic capacity degradation model; performing capacity prediction using trained stochastic capacity degradation model to obtain remaining useful life of one or more testing lithium-ion batteries; generating an air mass flow rate and a charging/discharging rate by a controller; and inputting the air mass flow rate and the charging/discharging rate into a battery thermal management system to improve battery RULs by adjusting lithium-ion battery temperature.

First claim

What is claimed is:

1 . A self-adaptive lithium-ion battery method using knowledge-reinforced machine learning (KRML) and Kalman filtering, comprising:
training an artificial neural network (ANN) and synchronizing the ANN with dual extended Kalman filters (DEKFs) to capture battery capacity data of each of one or more lithium-ion batteries; integrating prior knowledge with Gaussian process regression to form an integrated knowledge-reinforced Gaussian process regression; training a stochastic capacity degradation model by employing the integrated knowledge-reinforced Gaussian process regression with the captured battery capacity data to obtain a trained stochastic capacity degradation model; performing capacity prediction using the trained stochastic capacity degradation model to obtain remaining useful life (RUL) of one or more testing lithium-ion batteries; generating an air mass flow rate and a charging/discharging rate by a controller according to the RUL and a working condition of the one or more testing lithium-ion batteries; and inputting the air mass flow rate and the charging/discharging rate generated by the controller into a battery thermal management system (BTMS) to adjust lithium-ion battery temperature.

2 . The method according to claim 1 , wherein:
the prior knowledge includes that a battery capacity is a positive value and lower than a battery maximum capacity, and monotonically decreases over time.

3 . The method according to claim 1 , wherein:
the DEKFs includes a top EKF and a bottom EKF which are connected in parallel with each other.

4 . The method according to claim 1 , wherein:
the ANN is trained using a historical dataset from a baseline battery to capture dynamics of the baseline battery.

5 . The method according to claim 1 , wherein:
the KRML includes a diagnosis module using the ANN and the DEKFs, and a prognosis module using the integrated knowledge-reinforced Gaussian process regression.

6 . The method according to claim 1 , wherein:
the working condition of the one or more testing lithium-ion batteries includes a temperature, a charging/discharging profile, and/or a humidity.

7 . An electronic device, comprising:
a memory, configured to store program instructions for performing a self-adaptive lithium-ion battery method using knowledge-reinforced machine learning (KRML) and Kalman filtering; and a processor, coupled with the memory and, when executing the program instructions, configured for: training an artificial neural network (ANN) and synchronizing the ANN with dual extended Kalman filters (DEKFs) to capture battery capacity data of each of one or more lithium-ion batteries; integrating prior knowledge with Gaussian process regression to form an integrated knowledge-reinforced Gaussian process regression; training a stochastic capacity degradation model by employing the integrated knowledge-reinforced Gaussian process regression with the captured battery capacity data to obtain a trained stochastic capacity degradation model; performing capacity prediction using the trained stochastic capacity degradation model to obtain remaining useful life (RUL) of one or more testing lithium-ion batteries; generating an air mass flow rate and a charging/discharging rate by a controller according to the RUL and a working condition of the one or more testing lithium-ion batteries; and inputting the air mass flow rate and the charging/discharging rate generated by the controller into a battery thermal management system (BTMS) to adjust lithium-ion battery temperature.

8 . The electronic device according to claim 7 , wherein:
the prior knowledge includes that a battery capacity is a positive value and lower than a battery maximum capacity, and monotonically decreases over time.

9 . The electronic device according to claim 7 , wherein:
the DEKFs includes a top EKF and a bottom EKF which are connected in parallel with each other.

10 . The electronic device according to claim 7 , wherein:
the ANN is trained using a historical dataset from a baseline battery to capture dynamics of the baseline battery.

11 . The electronic device according to claim 7 , wherein:
the KRML includes a diagnosis module using the ANN and the DEKFs, and a prognosis module using the integrated knowledge-reinforced Gaussian process regression.

12 . The electronic device according to claim 7 , wherein:
the working condition of the one or more testing lithium-ion batteries includes a temperature, a charging/discharging profile, and/or a humidity.

13 . A non-transitory computer-readable storage medium, containing program instructions for, when being executed by a processor, performing a self-adaptive lithium-ion battery method using knowledge-reinforced machine learning (KRML) and Kalman filtering, the method comprising:
training an artificial neural network (ANN) and synchronizing the ANN with dual extended Kalman filters (DEKFs) to capture battery capacity data of each of one or more lithium-ion batteries; integrating prior knowledge with Gaussian process regression to form an integrated knowledge-reinforced Gaussian process regression; training a stochastic capacity degradation model by employing the integrated knowledge-reinforced Gaussian process regression with the captured battery capacity data to obtain a trained stochastic capacity degradation model; performing capacity prediction using the trained stochastic capacity degradation model to obtain remaining useful life (RUL) of one or more testing lithium-ion batteries; generating an air mass flow rate and a charging/discharging rate by a controller according to the RUL and a working condition of the one or more testing lithium-ion batteries; and inputting the air mass flow rate and the charging/discharging rate generated by the controller into a battery thermal management system (BTMS) to adjust lithium-ion battery temperature.

14 . The storage medium according to claim 13 , wherein:
the prior knowledge includes that a battery capacity is a positive value and lower than a battery maximum capacity, and monotonically decreases over time.

15 . The storage medium according to claim 13 , wherein:
the DEKFs includes a top EKF and a bottom EKF which are connected in parallel with each other.

16 . The storage medium according to claim 13 , wherein:
the ANN is trained using a historical dataset from a baseline battery to capture dynamics of the baseline battery.

17 . The storage medium according to claim 13 , wherein:
the KRML includes a diagnosis module using the ANN and the DEKFs, and a prognosis module using the integrated knowledge-reinforced Gaussian process regression.

18 . The storage medium according to claim 13 , wherein:
the working condition of the one or more testing lithium-ion batteries includes a temperature, a charging/discharging profile, and/or a humidity.

Assignees

• Scibot Tech Llc
• Univ Michigan Tech

Inventors

• Wang Zequn
• Bai Guangxing
• Li Lun

Classifications

• G06N3/08 Primary

  Learning methods · CPC title

• G05B13/027 Primary

  using neural networks only · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• Y02E60/10

  Energy storage using batteries · CPC title