Prediction of Plug-in Electric Vehicle's State-of-Charge using Gradient Boosting Method and Random Forest Method

The accurate prediction of state-of-charge (SOC) of plug-in electric vehicle (PEV) is a vital factor in order to decide the mode of operation. Depending upon SOC status at the trip end of PEVs, the operator decides the charging and discharging of vehicles. Therefore, accurate prediction is the need of the demand to know the exact status of PEVs SOC. The total trip distance of PEVs and energy consumption pattern of different sizes of PEVs are also included so that calculation of SOC can be made to accurate. The prediction analysis is carried out by using machine learning approaches like gradient boosting method (GBM) and random forest method (RFM). Tree based ensemble method improves the worst performance by adding additional trees. The additional trees in GBM further improve the prediction accuracy of PEVs SOC. The application of both the machine learning approaches creates significant impact on PEVs SOC prediction. Moreover, a comparative analysis is carried out between both the approaches and results have been presented. Three case studies with different sizes are considered to analyze the results of the proposed work.