Statistical Learning for Accurate and Interpretable Battery Lifetime Prediction

Abstract

Data-driven methods for battery lifetime prediction are attracting increasing attention for applications in which the degradation mechanisms are poorly understood and suitable training sets are available. However, while advanced machine learning and deep learning methods promise high performance with minimal data preprocessing, simpler linear models with engineered features often achieve comparable performance, especially for small training sets, while also providing physical and statistical interpretability. In this work, we use a previously published dataset to develop simple, accurate, and interpretable data-driven models for battery lifetime prediction. We first present the “capacity matrix” concept as a compact representation of battery electrochemical cycling data, along with a series of feature representations. We then create a number of univariate and multivariate models, many of which achieve comparable performance to the highest-performing models previously published for this dataset; thus, our work can serve as a comprehensive benchmarking study for this dataset. These models also provide insights into the degradation of these cells. Our approaches can be used both to quickly train models for a new battery cycling dataset and to benchmark the performance of more advanced machine learning methods.

Residual Life Prediction of Lithium Batteries Based on Data Mining

“…The grey data mining system draws on the structure of the traditional data mining system and makes full use of the grey system method to carry out data mining. It is based on a database and data warehouse and provides effective methods for data mining tools of data warehouse [ 16 ]. Figure 1 shows the system architecture.…”

“…When the battery capacity decreases to a certain critical point (typically 80% of the rated capacity, but different battery types have different thresholds), the battery is considered to be invalid [10]. Remaining service life refers to the remaining service life of a battery after it has been used for a period of time [11][12][13][14][15][16]. For example, the power lithium battery has a cycle life of 500 times; that is, it can last 500 times under normal charging and discharging conditions.…”

An open access tool for exploring machine learning model choice for battery life cycle prediction

“…A variety of machine learning and data-assisted techniques have been employed for battery lifetime prediction, with published predictive performance achieving error rates under 10% and as low as 0.2% (Tseng et al, 2015) for the best models. The least complex models are feature-based linear regression models which have the benefit of low computational effort and clear interpretability (Long et al, 2013;Xing et al, 2013;Tseng et al, 2015;Berecibar et al, 2016;Song et al, 2017;Severson et al, 2019;Chen et al, 2020;Attia et al, 2021;Gasper et al, 2021). Simple feature-based models also limit the potential for overfitting small training datasets (Sendek et al, 2022) but require the design and selection of features tailored to a given battery dataset.…”

“…The Variance and Discharge model are both implemented using a Scikit-learn pipeline which includes a StandardScaler step followed by an ElasticNetCV model. This model is implemented with 5-fold cross validation and L1 ratios of [0.1, 0.5, 0.7, 0.9, 0.95, 0.99, 1]-these values were referenced from Attia et al's (Attia et al, 2021)…”

A two-stage deep learning framework for early-stage lifetime prediction for lithium-ion batteries with consideration of features from multiple cycles

“…To represent the features' evolution curve, Paulson et aluse three multi-cycle features to capture the median, derivative with respect to cycles (Paulson et al, 2022). Attia et al (2021) extract the feature from the difference between 10th and 100th cycle along with fourteen summary statistic functions and four feature transformations. However, the aging process is a continuous process, and information can be extracted from the development trend of physics-guided features.…”