Multi-Module Model Incorporating Temporal Convolutional Networks for Enhanced Lithium-Ion Battery Capacity Estimation

With the advancement of new energy and energy storage technologies, the demands for enhanced performance and safety of lithium-ion batteries have grown, while traditional time series modeling methods struggle to balance local details with global trends. To address this issue, this study introduces an innovative multi-module framework called MTMH. Inspired by the block design of the Mamba model, MTMH captures fine-grained features across different time scales through a layered approach and employs a forgetting mechanism to eliminate ineffective historical information, thereby improving model stability and prediction precision. Moreover, the proposed architecture innovatively integrates temporal convolutional networks (TCNs) and multi-head self-attention (MHSA) mechanisms, which respectively capture local short-term dependencies and global long-distance relationships, synergistically enhancing the representation capability of complex time series data. The experimental outcomes reveal that MTMH outperforms numerous cutting-edge techniques in terms of precision and reliability for battery health forecasting, underscoring its potential for widespread use.