A Digital Twin-Driven Life Prediction Method of Lithium-Ion Batteries Based on Adaptive Model Evolution

Area of Science:

• Battery technology
• Reliability engineering
• Data science

Background:

• Accurate life prediction and reliability evaluation are crucial for lithium-ion battery predictive maintenance.
• Characterizing the dynamic and stochastic nature of battery life throughout its cycle is a persistent challenge.

Purpose of the Study:

• To propose a digital twin for reliability in lithium-ion batteries.
• To enhance remaining useful cycle life prediction and overall battery reliability assessment.

Main Methods:

• Developed capacity degradation and stochastic degradation models.
• Implemented life prediction and reliability evaluation models.
• Utilized a Bayesian algorithm for adaptive evolution of the digital twin model.
• Conducted experimental verification.

Main Results:

• The digital twin demonstrated high accuracy across the entire battery life cycle.
• The adaptive evolution algorithm maintained prediction error within approximately 5%.
• Significant reductions in predictive maintenance costs were observed (62.0% for L1, 52.5% for L6).

Conclusions:

• The proposed digital twin for reliability offers accurate life prediction and evaluation for lithium-ion batteries.
• Adaptive model evolution is key to achieving high prediction accuracy.
• The digital twin approach effectively supports predictive maintenance strategies, leading to substantial cost savings.