Battery performance and longevity are determined by SoC and SoH (state of charge and state of health). Therefore, SoC pertains to available charge compared to the battery’s total capacity, which varies with cycles in charging and discharging. On the other hand, SoH also measures the battery’s capacity to store and deliver the energy as compared to their own starting point, subject to the effects of the cycle aging, temperature effects, and chemical degradation. These metrics need to be estimated with sufficient accuracy and are important when managing energy for applications such as EVs and grid storage. Predictive modeling of SoC and SoH is carried out using long short-term memory (LSTM) networks on historical battery data as this study does. These models were trained using processed data of key parameters such as terminal voltage, current, temperature, and cycle. These were validated with robust validation scores of mean absolute error (MAE), root mean squared error (RMSE), and R2 values. The obtained results showed high predictive accuracy as SoC and SoH estimates were close to the actual values with a low error and an R2 score near 1. The research also analyzed power demand and instantaneous power and derived battery energy efficiency and operational behavior. The study proposes the state of functionality (SoF) metric that combines SoC, SoH, and power measures to determine the overall battery performance dynamically. The findings support the use of LSTM models for predictive battery management in the real-time monitoring as well as optimized charging strategy and better lifecycle assessment. Model accuracy will be refined, and battery technologies will be extended to other battery technologies in future research.