This project includes code and battery data related to Deng Z, Hu X, Xie Y, Xu L, Li P, Lin X, et al. Battery health evaluation using a short random segment of constant current charging. iScience. 2022:104260. https://doi.org/10.1016/j.isci.2022.104260.
In this study, four types of batteries are investigated, namely, LiNixCoyAl1−x−yO2 (NCA), LiNixMnyCo1−x−yO2 (NMC), LiFePO4 (LFP), and a blend of NMC and LiCoO2 (NMC-LCO). The dataset of NMC-LCO is contributed by Hawaii Natural Energy Institute (HNEI), and the other three are contributed by Sandia National Laboratories (SNL). The raw datasets of batteries come from BatteryArchive, which is a website to present battery data in a uniform format. People can download the battery datasets from the website or find it in the Release of this project.
In the Release of this project, HNEI_raw_data.mat and SNL_raw_data.mat are the raw data extracted from .cvs files, HNEI_cell.mat, LFP_cell.mat, NCA_cell.mat and NMC_cell.mat include battery charging data and capacity of each cycle. All the above data can be extracted by running extract_data.m
Correlations between battery state of health (SOH) and features are analysed. correlation_analysis_12seg.m is used to analyse correlation at a fixed number of segments, e.g. 12 in this study. correlation_analysis_diff_seg.m is used to analyse correlation at different numbers of segments in this study.
soh_MLR.m, soh_SGPR.m, and soh_DCNN.m are used to estimate battery SOH by using multiple linear regression (MLR), sparse Gaussian process regression (SGPR), and deep convolutional neural network (DCNN).
Matlab>=2018a
Gaussian Process Regression toolbox in http://www.GaussianProcess.org/gpml/code