Optimization Of Electric Vehicle Li-Battery Fuel Gauge Based On Machine Learning Algorithm

With people paying more and more attention to non-renewable energy resource and environment,lithium battery has been used to be the power of most electronic vehicle because it is environment-friendly battery with high energy density and high average output voltage.But because of it's own defect,every lithium battery electric source need a battery management system(BMS)to manage it's use and ensure safety.In a BMS,for many key technology,such as active balancing,protection of over charge and over discharge,quick charge and so on,batteries' real time state of charge(SOC)is an important parameter.The real time estimation of SOC plays an essential role in effective energy management.In many existing SOC estimation algorithm,some algorithm based on machine learning achieve a good accuracy.But for the situation of electric vehicle,current change acutely during car going.It would bring a bad result for these algorithm.This paper put forward two methods to improve this.Improved methods including a new training set for the model and voltage correction algorithm based on BP ANN.Finally,above two new methods is verified by a set of experimental data.Result shows that a new training set can make the root mean square error(RMSE)be below 2.5%,and max error below 20%.At the last,voltage correction bring a new decrease of RMSE below 1.9%,and max error below 7.6%.It proved above two methods could enhance both universal and accuracy of SOC estimation algorithm based on machine learning.