The lithium-ion battery industry is embracing unprecedented opportunities for growth. However, the lack of understanding for dynamic evolution of electrode microstructure during cycling, and the strong cross-scale interactions between microstructure and macroscopic properties, pose considerable challenges to the study of battery failure mechanisms. In this presentation, we will construct electrochemical cross-scale models (both for active particles and for the cell), making use of homogenization methods and deep learning techniques, to explore failure mechanisms of lithium batteries and to predict their electrochemical performance. The models will help to elucidate the impact mechanisms of the heterogeneous and dynamically evolving microstructures of porous electrodes on battery performance and provide optimization strategies for the design of electrode microstructures.