Robust State of Health estimation of lithium-ion batteries using convolutional neural network and random forest

The State of Health (SOH) lithium-ion batteries is directly related to their safety and efficiency, yet effective assessment SOH remains challenging for real-world applications. In this paper, the estimation (i.e., capacity fading) under partial discharge with different initial final Charge (SOC) levels investigated. challenge lies in fact that causes truncation data available estimation, thereby leading loss or distortion common indicators. To address challenge, we utilize convolutional neural network (CNN) extract indicators both changes ( Δ SOH) between two successive charge/discharge cycles. random forest algorithm then adopted produce estimate by exploiting from CNNs. Performance evaluation conducted using SOC ranges created a fast-discharging dataset. proposed approach compared (i) differential-analysis-based (ii) CNN-based approaches only indicators, respectively. Through comparison, demonstrates improved accuracy robustness. Sensitivity analysis CNN models further validates makes better use data. • Partial may distort cause capacity-related Convolutional nets can effectively discharge. Capacity difference consecutive cycles be used estimation. Indicators complement those capacity. Random fuse