This dissertation investigates supercapacitor performance and longevity prediction in electric vehicles through numerical modeling and machine learning approaches. The research addresses two key challenges: understanding self-discharge mechanisms and predicting remaining useful life. The first phase develops a numerical model analyzing supercapacitor self-discharge behavior, incorporating ion redistribution, charge transfer kinetics, side reactions, and temperature effects. Through computational fluid dynamics and electrochemical simulations, the study reveals the two-stage nature of self-discharge processes and identifies critical factors affecting voltage decay. The second phase uses deep learning to predict remaining useful life of energy storage systems. Using historical performance data and operational parameters, the neural network model employs Transformer and LSTM architectures to capture spatial and temporal degradation patterns. Results show the combined approach provides superior predictive capabilities. The numerical model achieves >90% correlation with experimental data, while the deep learning model demonstrates 92% accuracy for remaining useful life estimation. This research contributes: (1) comprehensive understanding of supercapacitor self-discharge mechanisms, (2) a validated numerical performance prediction model, and (3) an innovative deep learning framework for lifetime estimation. These findings directly improve reliability and efficiency of electric vehicle energy storage systems.