2022 4th International Conference on Energy, Power and Environment (ICEPE) · 2022 · 13 citations · 11 references
EngineeringComputer ScienceElectrical EngineeringDemand ForecastingForecastingMachine LearningCharging InfrastructureCharging SystemsEnergy Demand ForecastingLoad ForecastingState Of Charge EstimationSupport Vector MachinesCharging Demand ForecastingElectric Load ForecastingMachine Learning For State Of Charge EstimationMachine Learning Forecasting ModelsMachine Learning-based ForecastingState Of Charge ForecastingState Of Charge PredictionTraining Time
Electric vehicle (EV) is approaching to the market in huge scale. Futuristic planning is necessary to accommodate the EVs in changing station as well as in electrical system. Therefore, the possible charging demand by the EVs should be predicted for proper capable infrastructure. Machine Learning (ML) is the best technique for prediction. This current paper presents a comparative study among three ML techniques to determine the ideal ML technique to predict required SOC of EVs. Existing vehicle flow of Agartala city is used to train the ML model. The results show that, among CG-SVM, bagged tree regressor, and boosted tree regressor; CG-SVM outperforms the other two ML techniques with respect to performance metrics, such as-MSE, MAE, R-squared value and RMSE also been observed. CG-SVM has lesser training time and higher prediction speed. Thereby, establishing CG-SVM as suitable ML technique to predict the required SOC of EVs.
11