Performance Comparison of Conventional and Intelligent method of Charge Estimation

This paper focuses on the implementation and performance comparison of a conventional and an intelligent method for estimation of SoC of a battery. Two different methods of estimation have been selected after careful study and literature review. The first method is Linear Kalman Filter (LKF), which is a conventional method, widely in use. The second method selected is Neural network using Feed Forward. The final results of both the methods are compared and studied to draw a conclusion. Both the methods have been implemented in MATLAB software. For Kalman Filter implementation, Thevenin circuit is modelled to achieve the needed equations. These equations are used to calculate the predict the error which the updates the Kalman gain. In Neural networks, the implementation comprises of training and testing. Mini batches have been taken for the training of the network along with Adam optimizer.