Battery State of Charge Estimation Based on the Parameters Analysis and Machine Learning

Abstract

The appropriate evaluation of the charge state of the battery is critical for ensuring safety and avoiding potential malfunctions in electric cars, cell phones, computers, and medical devices. The battery management system, on the other hand, provides important functions such as guaranteeing safe operation and informing the user about the battery’s state. Several approaches for estimating the battery state of charge (SoC) have been presented to produce the most effective management system. Machine learning techniques are used to compare prediction accuracy and the region of the convergence curve on testing and training data in this study. In this project, we proposed the use of machine learning algorithms as a means of predicting the battery state of charge in several locations that use battery management systems. The suggested methodology states using Weka software to implement three different algorithms of machine learning on battery parameters such as constant and current, temperature, and voltage. Results indicate that the best structure obtained using Weka is the Random Forest having the maximum correlation coefficient by finding the root mean square error (RMSE). Contradictory, the three machine learning algorithms which are decision tree, random forest, and linear regression revealed that decision trees have low correlation and relatively high root mean square error. The significance of the present project relies on its ability to predict the state of charge, a necessary prerequisite to executing a sustainable battery management system in electrical grids, assisting operators in efficiently managing general power, and helping achieve energy efficiency and production to its maximum ability, we used the stander USB while implementing the prototype.