Abstract
<jats:p>Accurate prediction of remaining useful life (RUL) for lithium-ion batteries is critical for ensuring reliability and safety in energy storage systems. This paper introduces a novel hybrid deep learning architecture that synergistically combines bidirectional gated recurrent units with multi-head attention mechanisms and MLP-Mixer modules within a mixture of experts framework. The proposed model effectively captures complex temporal dependencies and degradation patterns from battery capacity time series data. Extensive experimental evaluation on NASA and CALCE public datasets demonstrates the superior performance of our approach, achieving relative errors of 0.005 and 0.0019 respectively, substantially outperforming existing state-of-the-art methods. The hybrid architecture provides robust RUL predictions while maintaining computational efficiency, offering significant improvements for battery health monitoring and prognostic applications.</jats:p>