Lithium-Ion Battery State-of-Charge and State-of-Energy Simultaneous Estimation Via Sparse- Quasi Recurrent Neural Networks(S-QRNN)

This paper presents an innovative methodology for the concurrent estimation of Lithium-Ion Battery (LiB) State-of-Charge (SoC) and State-of-Energy (SoE) employing Sparse Quasi-Recurrent Neural Networks (S-QRNN). The proposed framework is designed to leverage sparse connectivity patterns to efficiently capture intricate long-term dependencies within battery dynamics. Unlike traditional recurrent neural networks and Convolutional Networks, S-QRNN allows for more effective handling of sequential data, making them well-suited for predicting battery behavior, which exhibits complex temporal dynamics. Furthermore, the sparse connectivity structure reduces computational complexity and enhances the interpretability of the model. To validate the effectiveness and accuracy, adequate experimentation was conducted using laboratory-produced battery data. Moreover, the accuracy and computational efficacy of the proposed scheme have been verified in an OPAL-RT-based Real-Time Power Hardware-In-Loop (PHIL) environment. The Opal RT platform provides a reliable and flexible environment integrated with MATLAB/Simulink for hardware-in-loop simulation. Experimental results demonstrate that the proposed method achieves robust and accurate estimation of both SoC and SoE, even in dynamic operational conditions of temperatures and battery load profiles.