RBBMO‐STOReAT‐net driven framework for effective SOC and SOH prediction in lithium‐ion batteries using electrochemical spectroscopy

Abstract

Accurately estimating the state of charge (SOC) and state of health (SOH) of lithium‐ion batteries (LIBs) is crucial for assuring safety and longevity in electric vehicles (EVs) and energy storage devices. However, existing data driven models frequently face issues like noise affected electrochemical impedance spectroscopy (EIS) measurements, and nonlinear degradation behaviour. Although synthetic labels enable the development and benchmarking of SOC‐SOH estimation frameworks when direct lifetime measurements are unavailable, they introduce potential bias and may overestimate predictive performance. The core innovation of the proposed Red Billed Blue Magpie Optimization‐Stochastic Latent Transformer Reservoir Computing Attention Network (RBBMO‐STOReAT‐Net) framework lies in the integration of three complementary components that collectively address the major challenges of SOC and SOH estimation from EIS data. The framework includes red billed blue magpie optimization (RBBMO) based stochastic latent transformer (SLT) with reservoir computing attention network (RCAN) for SOC‐SOH estimation. The hybrid DL model integrates the latent feature extraction and nonlinear temporal representation capabilities of the RCAN and SLT modules. The RBBMO technique is employed for optimizing hyperparameter globally and prevents premature convergence. Experimental validation using EIS of lithium‐ion batteries dataset with Python software was carried out, illustrating the superior precision on both SOC‐SOH prediction. For SOC, the model achieves R ² score of 0.9994, mean absolute percentage error (MAPE), mean squared error (MSE), median absolute error (median AE), mean absolute error (MAE), root mean square error (RMSE), and max error of 0.03530, 0.00064, 0.01437, 0.01875, 0.02528, and 0.12916. Similarly, SOH attains a R ² Score of 0.9992, MAPE, MSE, median AE, MAE, RMSE, and max error values of 0.04505, 0.00084, 0.02145, 0.02394, 0.02903, and 0.09787. Results confirm the proposed RBBMO‐tuned STOReAT‐Net outperforms state‐of‐the‐art methods in battery management systems.