State of Health estimation for lithium-ion batteries using Random Forest and Gated Recurrent Unit

Lithium-ion batteries are widely used in electric vehicles, energy storage and other fields, and the State of Health ( SOH ) estimation of lithium-ion batteries are key to ensure the safe operation of battery systems. In this paper, a method combining Empirical Modal Decomposition (EMD), Random Forest (RF) and Gated Recurrent Unit (GRU) for SOH estimation of lithium-ion batteries was proposed. In this approach, we first extracted the time interval during equal voltage increase and the time interval of equal voltage decrease as health indicators (HIs), and analyzed the correlation between the health indicators and SOH using Pearson's coefficient. After that, the empirical modal decomposition (EMD) was used to decompose the battery SOH data, and the Variance Contribution Ratio (VCR) was introduced to measure the relationship between the intrinsic modal function (imf) component and SOH. Finally, an EMD-VCR-GRU-RF based SOH estimation model was developed. The prediction results show that the EMD-VCR-GRU-RF model has the smallest prediction error and the model computation time is at least 15.84 % less than that of the EMD-VCR-GRU model. Our work effectively applies deep learning and machine learning to battery health management, balancing prediction accuracy and computational efficiency. It provides support and reference for battery health management and smart operation and maintenance . • Extracted easily measurable health indicators for battery charging and discharging • A method for EMD decomposition of battery SOH data is proposed. • Introducing the VCR to measure the weight of the imf component on SOH • A prediction model combining machine learning and deep learning is developed. • The proposed model saves at least 15.84 % of prediction time.