Advanced SOC Prediction for Lithium-Ion Batteries Using FNN Machine Learning Techniques: A Bayesian Regularization Training Approach

Abstract

The health of the battery can be examined by researchers as an element for the useful life of the lithium-ion battery and this is crucial for BMS. For these reasons, the goal of many researchers is to monitor the lithium-ion battery through the State of Charge. SOC Analysis is crucial for predicting the End of Life for LIB. Many Machine Learning techniques have tried to calculate the phenomenon of cell degradation, but defective algorithms can lead to distorted results, overfitting or underfitting, and excessive computational burdens. In this article, we have chosen to use a sophisticated neural network (FFN) technique for the predictive calculation of SOC, with an evaluation of the best performance in terms of RMSE - Root Mean Squared Error. The new algorithm is developed with the use of Matlab software, simulating Datasets of NASA PCoE Research Center, based on the comparison of other Machine Learning models. Finally, the analysis of the Actual SOC and the Predicted SOC curves is used to represent the precision of the proposed algorithm and predict the degradation phenomenon of the LIB. In the future, the author would like to test the algorithm to look for a lithium-ion battery that will last forever.