Lithium-ion batteries (LIBs) play a crucial role for various energy storage applications; however, early-life prediction remains challenging. This paper proposes a Multi-Channel Stacked Graph Convolutional Neural Network (MNGCN) model, which constructs a multi-channel stacked graph structure to capture the complex relationships among battery features, and utilizes Graph Convolutional Networks (GCN) to extract key features. Additionally, the non-Dominated Sorting Genetic Algorithm II (NSGA-II) is applied to optimize the model parameters, enhancing prediction accuracy. Experiments were conducted on the lithium-ion battery dataset released by MIT and compared with several benchmark models. The results show that the proposed MNGCN model outperforms traditional models across multiple performance metrics, including RMSE, MAE, R2, and SDEX, significantly enhancing prediction accuracy and model robustness. The MNGCN not only demonstrates advantages in capturing complex feature interactions and nonlinear relationships but also leverages the optimization algorithm to better extract latent information from the data, achieving more accurate and stable battery life predictions.