This paper proposes an advanced approach for the Lithium-ion (Li-ion) battery state of health (SOH) estimation using a Two Pulse Load Test (TPLT) and a combination of multi-step machine learning models (MLMs), a Symbolic Regression (SR), Gaussian process regression (GPR) and neural networks (NN). This proposed system offers a robust and lightweight method for SOH estimation, making it well-suited for real-time applications, particularly in resource constrained embedded Battery Management System (BMS) environments. The model is validated through real and simulated TPLT data from LG Chem MH1 18650 and Panasonic NCR18650PF cells, achieving an average root mean squared error (RMSE) of 1.06%. Implementation of the proposed method on a Raspberry Pi 4 further demonstrates its high feasibility for real-time operation.
