Gaussian Process Regression With Automatic Relevance Determination Kernel for Calendar Aging Prediction of Lithium-Ion Batteries

Abstract

Battery calendar aging prediction is of extreme importance for developing durable electric vehicles. This paper derives machine learning-enabled calendar aging prediction for lithium-ion batteries. Specifically, the Gaussian process regression (GPR) technique is employed to capture the underlying mapping among capacity, storage temperature, and SOC. By modifying the isotropic kernel function with an automatic relevance determination (ARD) structure, high relevant input features can be effectively extracted to improve prediction accuracy and robustness. Experimental battery calendar aging data from nine storage cases are utilized for model training, validation, and comparison, which is more meaningful and practical than using the data from a single condition. Illustrative results demonstrate that the proposed GPR model with ARD Matern32 (M32) kernel outperforms other counterparts and can achieve reliable prediction results for all storage cases. Even for the partial-data training test, multi-step prediction test and accelerated aging training test, the proposed ARDbased GPR model is still capable of excavating the useful features, therefore offering good generalization ability and accurate prediction results for calendar aging under various storage conditions. This is the first known data-driven application that utilizes the GPR with ARD kernel to perform battery calendar aging prognosis.Index Terms-Lithium-ion batteries, calendar aging prediction, battery health, Gaussian process regression, datadriven model.

State-of-Health Prognosis for Lithium-Ion Batteries Considering the Limitations in Measurements via Maximal Information Entropy and Collective Sparse Variational Gaussian Process

“…A hybrid GP function regression was proposed to overcome the deteriorations in trends when the test data is multiple steps ahead of the training ones [50]. A GPR with automatic relevance determination kernels, which can extract high relevant features to enhance the robustness and prediction accuracy, was devised to prognosticate lithium-ion battery's calendar aging [51]. However, traditional GPR is constrained by its computational and storage complexity, while sparse variation inference offers a convenient way out [20].…”

“…Usually, SOH estimation metrics are categorized into two kinds: model-based methods [13], [40]- [44] and data-driven methods [14]- [19], [21]- [24], [31], [45]- [51]. Admittedly, the current SOH prognostics methods have already realized relatively satisfying results.…”

A Facile Approach to High Precision Detection of Cell-to-Cell Variation for Li-ion Batteries

“…Considering the correlation between aging speed and internal chemistry, it is reasonable that cells with similar chemical/physical states at certain time point will experience different aging speeds under the same environmental stress (such as electric stress, temperature stress, mechanical stress) and present varied chemical/physical states in the following time points 18,42,43 . In this study, cell degradation under room temperature storage was mainly caused by the SEI film formation side reaction, which is influenced by the surface area of the anode active material, the thickness of the anode and the original SEI film formed during the preparation process 44,45 . Therefore, although the cells with similar chemical/physical states were degraded under the same operation conditions, they will still experience different aging speeds over a long time period.…”

Model Migration Neural Network for Predicting Battery Aging Trajectories

“…However, the parameters in these simplified models usually lack physical meanings. It would be difficult to add reasonable constraints during the identification process to prevent potential over-fitting [29]. As a result, this kind of methods would be sensitive to the noise and easy to diverge in the prediction phase [30].…”