A smoothing-assisted machine learning framework improves the accuracy of state-of-health predictions across the lifespan of lithium-ion batteries, with notable performance gains observed during the late stages of cycling.
Lithium-ion batteries are widely used in electric vehicles and energy storage systems because of their high energy density, long cycle life, and fast charging capability. Over time, however, battery performance gradually degrades due to electrochemical aging, thermal effects, mechanical stress, and varying operating conditions. This degradation leads to progressive capacity loss and an overall reduction in State of Health (SOH). Accurate SOH prediction is therefore essential for safe battery operation and effective battery management system (BMS) control.
Although machine learning techniques are capable of modeling complex nonlinear battery degradation, their prediction accuracy often decreases substantially during the late stages of battery cycling. As cycling advances into these final operating phases, degradation behavior becomes significantly more irregular while noise measurement increases.
To overcome this limitation, the authors proposed a smoothing-assisted SOH prediction framework utilizing a Random Forest regression model combined with rolling mean and Savitzky–Golay smoothing techniques. These dual smoothing methods effectively reduce short-term measurement noise while systematically preserving long-term degradation trends.
Experimental evaluation confirmed that the combined smoothing-assisted approach improves State of Health prediction accuracy across the entire battery lifespan.
Subject of Research: Machine learning state-of-health prediction for lithium-ion batteries using smoothing techniques
Article Title: Enhanced state-of-health prediction of lithium-ion batteries using machine learning with rolling mean and Savitzky–Golay smoothing techniques
Article References: Bala, S., Kanna, R., Gnanasekar, N., & Thangavelu, P. (2026). Enhanced state-of-health prediction of lithium-ion batteries using machine learning with rolling mean and Savitzky–Golay smoothing techniques. Ionics. https://doi.org/10.1007/s11581-026-07645-2
Image Credits: AI Generated
DOI: 10.1007/s11581-026-07645-2
Keywords: lithium-ion batteries, state-of-health prediction, battery management systems, machine learning, Random Forest regression, rolling mean, Savitzky-Golay smoothing, battery degradation, electrochemical aging, thermal effects, mechanical stress, capacity loss, energy storage systems, electric vehicles, late-stage cycling, noise reduction, nonlinear degradation
