Comparative Analysis of State of Charge Estimation Techniques for Electric Vehicle Batteries using various Artificial Neural Networks

State of Charge (SOC) estimation is a critical task in electric vehicle (EV) battery management systems, ensuring reliable and efficient energy management. However, the nonlinear behavior of Li-ion batteries, influenced by temperature, health, and SOC dependencies, poses significant challenges for accurate estimation. Traditional approaches, such as electrochemical models, require detailed knowledge of battery composition and physical responses, making them less adaptable to varying conditions. This paper proposes a comparative analysis of SOC estimation techniques using neural network models: Feedfor-ward Neural Network (FNN), Elman Recurrent Neural Network (ERNN), and Convolutional Neural Network (CNN). The models were optimized using the Adam optimizer, a 2-hidden-layer architecture, dropout layers, and trained for 60 epochs on a data set with voltage (V), current (I) and temperature (T) as input features. The data set spans a temperature range of -10°C to 25°C, with performance evaluated using RMSE, MSE and MAE metrics. The results indicate that, while FNN serves as a reliable baseline, ERNN outperforms it by capturing temporal dependencies in time-series data. CNN demonstrates the highest prediction accuracy, using its ability to extract complex features from sequential data. These findings provide valuable insights into the suitability of neural network models for SOC estimation and highlight CNN’s potential for advancing battery management systems in electric vehicles.