Recent advances in artificial intelligence (AI) have revolutionized the capabilities of battery management systems (BMS), particularly in the critical task of state of charge (SOC) estimation for lithium-ion batteries (LIBs). While conventional methods struggle with dynamic operating scenarios, our research introduces a novel deep neural network (DNN) architecture designed to handle the complexities of rapid charge–discharge cycles, and thermal fluctuations. We focused on developing a real-time monitoring solution that overcomes the traditional challenges of SOC estimation, where battery behavior exhibits nonlinear characteristics across varying operational parameters. Our methodology leverages a comprehensive dataset derived from four driving profiles: US06, LA92, Highway Fuel Economy Test (HWFET), and Urban Dynamometer Driving Schedule (UDDS). The experimental design incorporates an 80–20 split for training and testing, with a unique validation approach using randomly combined drive cycles to assess real-world adaptability. The DNN processes three key input parameters, current, voltage, and temperature, to generate an accurate SOC prediction. Comparative analysis against the second-order RC Equivalent Circuit Model (ECM) with Extended Kalman Filter (EKF) demonstrates the superior performance of our approach. Our implementation achieves remarkable accuracy with a Root Mean Square Error (RMSE) below 3.3% and Mean Absolute Error (MAE) under 2.6%, representing a significant advancement in SOC estimation technology.