Aging Trends and Predictive Value of Equivalent Circuit Parameters: The Role of Operating Conditions in Battery State-of-Health Estimation
- 1Department of Electrical Engineering, Kocaeli University, Kocaeli 41001, Türkiye
Abstract
This study investigates how two-RC equivalent circuit model (ECM) parameters reflect battery aging and contribute to state-of-health (SOH) estimation under different operating conditions. The analysis uses 12,643 measurements from 30 LG HE4 cells aged at 23–45 °C within five state-of-charge (SOC) windows spanning 0–100%. Parameters are identified from the first 300 s of post-discharge relaxation. Extreme gradient boosting (XGBoost) estimates SOH and quantifies predictive relevance through gain and SHapley Additive exPlanations (SHAP); Gaussian process regression (GPR) and Ridge regression assess dependence on model choice using the same cell-disjoint folds. Sampled fast-response resistance R0 shows the clearest aging trend, followed by the pulse-to-rest indicator Rpr; individual branch resistances and time constants vary more strongly with operating point. XGBoost mean absolute percentage error (MAPE) decreases from 2.014% for ECM-only inputs to 1.444% with temperature, 1.688% with SOC-window information, and 1.023% with both, with corresponding R2 values of 0.850, 0.931, 0.874, and 0.963. Temperature provides the larger separate improvement. R0 remains prominent across models, whereas other feature rankings depend on the learner and available condition information. Aging sensitivity and predictive relevance therefore provide complementary, condition-dependent views of battery health.
Related
- Two-Stage Chemical Degradation of PTFE with H- and HO-Radicals: A DFT Study Under PEMFC Operating Conditions
- Tuning the Selectivity of Glycerol Electrooxidation via Catalyst Composition and Operating Conditions
- Predictive-feedforward energy management for a hybrid agricultural machine operating in mountainous orchards
- Operating conditions redefine the governing principles of catalyst-layer fabrication
- Consistent Modeling of SEI Growth via Electron Diffusion Validated Across Diverse Cell Specifications and Operating Conditions in LFP/Graphite Cells
BibTeX
@article{kutal2026aging,
title = {{Aging Trends and Predictive Value of Equivalent Circuit Parameters: The Role of Operating Conditions in Battery State-of-Health Estimation}},
author = {Mücahit Kutal and Birol Arifoğlu},
journal = {Batteries},
year = {2026},
month = oct,
doi = {10.3390/batteries12110442},
url = {https://www.mdpi.com/2313-0105/12/11/442},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Aging Trends and Predictive Value of Equivalent Circuit Parameters: The Role of Operating Conditions in Battery State-of-Health Estimation
AU - Mücahit Kutal
AU - Birol Arifoğlu
PY - 2026
DA - 2026/10/23
JO - Batteries
DO - 10.3390/batteries12110442
UR - https://www.mdpi.com/2313-0105/12/11/442
AB - This study investigates how two-RC equivalent circuit model (ECM) parameters reflect battery aging and contribute to state-of-health (SOH) estimation under different operating conditions. The analysis uses 12,643 measurements from 30 LG HE4 cells aged at 23–45 °C within five state-of-charge (SOC) windows spanning 0–100%. Parameters are identified from the first 300 s of post-discharge relaxation. Extreme gradient boosting (XGBoost) estimates SOH and quantifies predictive relevance through gain and SHapley Additive exPlanations (SHAP); Gaussian process regression (GPR) and Ridge regression assess dependence on model choice using the same cell-disjoint folds. Sampled fast-response resistance R0 shows the clearest aging trend, followed by the pulse-to-rest indicator Rpr; individual branch resistances and time constants vary more strongly with operating point. XGBoost mean absolute percentage error (MAPE) decreases from 2.014% for ECM-only inputs to 1.444% with temperature, 1.688% with SOC-window information, and 1.023% with both, with corresponding R2 values of 0.850, 0.931, 0.874, and 0.963. Temperature provides the larger separate improvement. R0 remains prominent across models, whereas other feature rankings depend on the learner and available condition information. Aging sensitivity and predictive relevance therefore provide complementary, condition-dependent views of battery health.
SN - 2313-0105
ER -