Abstract:This study develops a methodology by capturing both the battery aging state and degradation rate for improved life prediction performance. The aging state is indicated by six physical features of an equivalent circuit model that are extracted from the voltage relaxation data. And the degradation rate is captured by two features extracted from the differences between the voltage relaxation curves within a moving window (for life prediction), or the differences between the capacity vs. voltage curves at different cycles (for life classification). Two machine learning models, which are constructed based on Gaussian Processes, are used to describe the relationships between these physical features and battery lifetimes for the life prediction and classification, respectively. The methodology is validated with the aging data of 74 battery cells of three different types. Experimental results show that based on only 3-12 minutes' sampling data, the method with novel features predicts accurate battery lifetimes, with the prediction accuracy improved by up to 67.09% compared with the benchmark method. And the batteries are classified into three groups (long, medium, and short) with an overall accuracy larger than 90% based on only two adjacent cycles' information, enabling the highly efficient regrouping of retired batteries.

What problem does this paper attempt to address?

The paper attempts to address the issue of reduced operational reliability caused by the aging of lithium-ion batteries during use, which makes the prediction of remaining useful life (RUL) a critical function in battery management systems. However, due to differences in aging characteristics and operating conditions among batteries, RUL prediction in practical applications faces significant challenges, and it also increases the difficulty of large-scale retired battery classification. Specifically, the study aims to improve in-situ life prediction and classification performance by capturing the aging state and degradation rate of batteries. The main objectives include: 1. **Improving the accuracy of battery life prediction**: By extracting features that reflect the current aging state and degradation rate of the battery, constructing machine learning models to more accurately predict the remaining useful life of the battery. 2. **Achieving efficient battery classification**: By using data from only two adjacent cycles, classifying batteries into three categories: long life, medium life, and short life, to facilitate the efficient reorganization of retired batteries. To achieve these goals, the study proposes a new method that combines two types of features: - **State features**: Six physical features of the equivalent circuit model (ECM) extracted from voltage relaxation data, used to indicate the aging state of the battery. - **Rate features**: Features extracted from the differences in voltage relaxation curves within a moving window or the differences in capacity-voltage curves of different cycles, used to capture the degradation rate of the battery. Two machine learning models constructed using Gaussian Process Regression (GPR) are used for life prediction and classification, respectively. Experimental results show that based on sampling data of only 3-12 minutes, the prediction accuracy of this method is improved by 67.09% compared to the benchmark method, and it can classify batteries with an overall accuracy of over 90% based on information from only two adjacent cycles.

Improve in-situ life prediction and classification performance by capturing both the present state and evolution rate of battery aging

Improving in-situ life prediction and classification performance by capturing both the present state and evolution rate of battery aging

A Physics-Informed Hybrid Multitask Learning for Lithium-Ion Battery Full-Life Aging Estimation at Early Lifetime

A Physics-Informed Hybrid Data-Driven Approach with Generative Electrode-Level Features for Lithium-Ion Battery Health Prognostics

Enhanced Gaussian Process Dynamical Modeling for Battery Health Status Forecasting

An Interpretable Semi-Supervised Learning Approach for Battery Lifespan Early Prognostic

Data-driven State-of-health Estimation for Lithium-Ion Battery Based on Aging Features

Accurate battery lifetime prediction across diverse aging conditions with deep learning

Predicting Battery Lifetime Under Varying Usage Conditions from Early Aging Data

A Data-Driven Approach for Lithium-ion Battery Lifetime Classification Based on Early Cycles

Lithium battery state-of-health estimation and remaining useful lifetime prediction based on non-parametric aging model and particle filter algorithm

A Data-Driven Method for Extracting Aging Features to Accurately Predict The Battery Health

Early prediction of lithium-ion battery lifetime via a hybrid deep learning model

Joint modeling for early predictions of Li-ion battery cycle life and degradation trajectory

Aging Characteristics-Based Health Diagnosis and Remaining Useful Life Prognostics for Lithium-Ion Batteries

A machine-learning prediction method of lithium-ion battery life based on charge process for different applications

Predicting Battery Aging Trajectory via a Migrated Aging Model and Bayesian Monte Carlo Method

Predictive Battery Health Management With Transfer Learning and Online Model Correction

A novel hybrid model for lithium-ion batteries lifespan prediction with high accuracy and interpretability

A deep learning approach incorporating attention mechanism and transfer learning for lithium-ion battery lifespan prediction