State of Charge Estimation by Using Artificial Neural Networks for Lithium Polymer Battery of Electric Vehicle

Battery State of Charge (SOC) estimation is of utmost importance for the efficient operation of battery systems. Artificial Neural Networks (ANN) have emerged as powerful tools for accurately estimating SOC. This study explores the application of ANN for SOC estimation in batteries using a comprehensive dataset obtained by simulating the usage of a lithium polymer cell model, specifically the ePLB C020, in an electric car resembling the Nissan Leaf. The dataset encompasses various battery parameters, including voltage, current, temperature, and other relevant variables. The research focuses on developing an optimized ANN architecture, training methodology, and evaluation metrics to ensure precise SOC estimation. The proposed ANN architecture consists of input, hidden, and output layers with carefully optimized neuron numbers and activation functions. Through an iterative training process employing backpropagation and gradient-based optimization algorithms, the weights and biases of the ANN are adjusted to enhance its performance. Evaluation metrics such as Mean Squared Error and correlation coefficient are employed to assess the accuracy and reliability of the SOC estimations. The experimental findings underscore the effectiveness of the ANN model in achieving accurate SOC estimation for the simulated electric car battery. This study highlights the potential practical applications of ANN in battery management systems, enabling reliable SOC estimation for improved control and optimization strategies. Additionally, a comparison between the proposed ANN model and Extreme Learning Machines (ELM) reveals superior performance of the ANN in SOC estimation for the simulated electric car battery.